Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Contents
Contents
II
Abstract
IV
Preface
VII
1. Introduction
1.1.1
1.1.2
7
8
2.1.1
2.1.2
10
2.1.3
11
11
2.2.1
12
2.2.2
14
15
2.3.1
Hierarchical Categorizations
16
2.3.2
Other Classifications
17
2.3.3
17
18
2.4.1
Extraversion
19
2.4.2
Agreeableness
20
2.4.3
Conscientiousness
21
2.4.4
Emotional Stability
21
2.4.5
22
23
2.5.1
23
2.5.2
24
II
3. Research Method
3.1 Making the Variables Operational in the Survey
26
26
3.1.1
27
3.1.2
29
3.1.3
30
30
32
32
35
4.2.1
36
4.2.2
Colouring the Picture: Who Are the People Behind the Components
38
4.2.3
Discussion
42
44
4.3.1
44
4.3.2
46
4.3.3
Discussion
48
51
51
52
55
6. References
57
6.1 Books/Articles
57
6.2 Websites
59
III
Abstract
MUDs (Multi-User Dungeons) are the predecessors of the graphical MMOGs (massively
multiplayer online games) or MMORPGs (adding RP for role-playing), but still continue to
draw in players despite the popularity of the latter. Bartle (1996;2003) has argued that people
play virtual worlds, because players want to have fun. What parts of the game bring fun to a
player differs per person and Bartle distinguished between four different player types: the
achiever, the explorer, the killer and the socialiser. Although these are widely known and
used, several authors criticise this categorization and therefore this research attempted to
test the typology by using empirical data. A fifth additional player type (the role-player) was
added, based upon the findings of Yee (2006a). Furthermore it was deemed that online
behaviour should have at least some root in offline personality, so the personality traits of the
Big Five (extraversion, agreeableness, conscientiousness, emotional stability and openness
to experience) were inserted in the online MUD player survey, held during May 2007, as well.
A principal component analysis was performed on the player type items and the four
initial player types were reproduced together with the additional role-player type. Linking the
components to demographics, MUD type, player characteristics and offline personality by
doing several multiple regression analyses, it was found that offline personality leads to a
significant improvement of the explained variance among the player types. Furthermore the
player types seemed to be much related to the characteristics of the main type of MUD that
they play. The findings were used to create a new model for player types. The different types
of game and their subsequent playing style at the bottom, the achiever player type above
that base, trying to advance within the specific MUD type and the explorer player type on top,
looking for new ways to extract fun from the game, when achievement has been gained. This
model has a lot in common with the hierarchical player categorization of Hedron (1998) and
links Bartle, Yee and Hedron into one model.
IV
Preface
With pride I present to you my master thesis about playing styles for online games. Although
finished in time for the deadline, it took some time to reach this state. Being a musician
myself, I was certain, even before the thesis themes of leisure studies were introduced, that I
wanted to graduate on something music related. However, the themes that would allow a
music related thesis did not match my main musical interest the personal enjoyment of
music itself. Fortunately, the theme virtual communities neatly provided another option,
because I am also administrator of such a virtual community: the Multi User Dungeon (MUD)
Rage of Vengeance.
The work on this theme lead to a theoretical paper and individual thesis proposal
about the consequences of playing at first, but this proved far too ambitious to be
conducted as a research in the given time frame. The player types of Bartle, which were
already somewhat included in the first theoretical paper, had fascinated me before and
therefore, when I had to make a choice between either the qualitative side on consequences
or the quantitative side for playing styles, I chose the latter, knowing that I was throwing away
almost ninety percent of my previous work. I never regretted making that choice though and I
enjoyed writing this thesis very much, although it was written in a tight time frame combined
with several other study related obligations and my own leisure activities.
One of the most important of these leisure activities, that received far too little
attention during the coming along of this project, is my own MUD home: Rage of Vengeance.
At this point I want to thank all of the players and administrators that currently play and have
played Rage of Vengeance in the past. Next to Rage of Vengeance, I of course want to
thank Erik van Ingen as my supervisor. Being the lazy and more leisure enjoying individual
that I am, he kept me going on this project, even when the time schedule became more and
more tight, especially because of the massive load of returned surveys that had to be put into
the statistical programme of SPSS. This leads directly to a very great amount of thanks that I
want to hand out to all players and MUD users that have taken the ten minutes time to fill out
my survey and all administrators that decided that this research was worth their attention and
put it through to their player bases.
VI
Admin(istrator)
Bugs
Builder
A MUD user that creates new areas or content for the MUD.
Code base
Coder
God
Immortal
MMOG
MMO(RP)Gs
MMORPG
MOO
MU*
MUD
MUDs
MUSH
Newbie
The overarching name for any player new to a specific MUD or MUDs
in general.
OOC
PK
Player
RL
RP
Wizard (Witch)
XP
VII
1. Introduction
MUDs are places where the imaginary meets the real (Bartle, 2003, p1). Insiders will
probably agree immediately with this statement, but for outsiders this statement probably
raises more questions than it provides an answer. The term MUD stands for Multi User
Dungeon and in essence it is a computer programme that people can log in to and immerse
themselves in an imaginary dungeon. Real people can interact with an online persisting,
imaginary environment and with several other real life persons around the world. This online
world is created solely by text. You are able to see where you are by reading descriptions
like: the immense wall that surrounds the castle is an impressive sight or a beautiful
sparkling fountain decorates this square. You handle your character by typing commands as
get sword and drink waterskin leading to textual responses as you get a sharp
broadsword or you drink water from a waterskin, and you can see other characters entering
or leaving the room that you are in and interact with them: Costrako sneaks in from the east
and you grin evilly at Costrako.
The first MUDs stem from the early days of the internet and were mainly designed to
be games. In essence, they had a lot in common with early text-based single player
adventure games like ADVENT, ZORK and HAUNT1, but with multiple players who could
communicate and cooperate with each other. The very first MUD, simply called MUD at the
time, was developed from 1978 onwards by Trubshaw and Bartle at Essex University,
England, and can still be found running under the name British Legends nowadays
(www.british-legends.com). Inspired by this MUD, later to be known as MUD1 to distinguish it
from the other MUDs, several new ones were developed, with various themes; from pure
player-killing MUDs (where the main goal is to kill other players) to more social role-playing
MUDs (in which the focus is on story and interaction) and even MUDs without any combat
system at all, purely based upon creating, building and inhabiting a virtual world. New code
1
Due to the computer systems at the time, these names had an all uppercase filename with a maximum of six characters. In
fact, the D for dungeon in MUD stems from a version of ZORK: DUNGEN. (Bartle, 2003)
bases were developed, giving rise to a full MUD tree (Keegan, 2003). This tree displays the
development of the genre and the rise of different MUDs. Most stem from an early version of
AberMUD which links back to MUD1.
Due to these developments in MUDs, the term MUD became loaded. Some MUD
players, particularly among social MUD users, felt a strong distinction between social MUDs,
where the focus lies on socialising with other players, and game-like MUDs, in which the
game is more profound. The term MUD was linked to game-like MUDs only, while the social
MUDs tried to find new names, leading to names like MOO (MUD Object Oriented) and
MUSH (Multi User Shared Habitat or Hallucination). This lead to a lot of ambiguity, leading to
a new overarching term, MU*, although this never became very popular2.
The genre has use for an overarching term, even more so because of the rise of the
graphical MUDs, and therefore this thesis will use the term MUD in its original meaning.
Graphical MUDs, often labelled MMORPGs (Massively Multiplayer Online Role Playing
Games) or more recently MMOGs, leaving out the role playing aspect, have many
similarities, but also some structural differences to text-based MUDs. Therefore this thesis
will use the term MUD to represent Multi User Dungeons involving multiple users inhabiting
an online, virtual and text-based world3. The users of MUDs will in this paper often be
referred to as players, since it is the easiest way of classification, and dates back to the early
MUD1 as a game.
Bartle (2003) claims that this is analogous to naming dinosaurs *saurs, because dino (or deinos in Greek, meaning fearful)
might imply that all saurs (or saurus, meaning lizard) are carnivorous monsters, while there are definitely other types as
well.
3
The acronym MUD has also been said to stand for multi user domain or dimension. This is all the same, as long as it indicates
ALL text based versions of MUDs.
MUDs in his book, although his main focus is on their sometimes dramatic impact on real
life4.
Virtual communities in general and MUDs in particular are an interesting topic for
research. Not only are virtual communities and online games a rather new phenomenon, it is
a massively growing one as well. For example the South-Korean Cyworld has twenty-two
billion page views per month (Oates, 2006) and, more specific on online games, Second Life
now has more than eight and a half million residents listed of which more than one and a
half million logged in within the last sixty days (www.secondlife.com, figures from August 5,
2007), while Blossom (2006) only mentioned slightly more than one million residents for
Second Life at the end of October 2006. These figures indicate that a lot of people seem to
be interested in being part of online communities and making contact with people that they
have never seen once in their lifetime and might at that never meet in the flesh. Still people
keep connecting and participating in communities, often several hours a week even when
they have a fulltime job. There must be something drawing them in.
In this context: Some authors mention MUD to be an acronym for Multi Undergraduate Destroyer due to the popularity among
and classes failed by students because of too much time spent on playing MUDs. (for instance by Reid, 1994)
His newer model stemming from 2003 has eight categories, but as he indicates himself: The conflicts between some of the
eight are meaningful, but for others the old types work just as well and are better at encapsulation. (Bartle, 2003, p.170)
the most points, or try to attain the highest rank in a guild, or plot against a more powerful
player to be able to take over his or her position. Most actions, strategies and goals for an
achiever will be related to achievement within the game context.
Although Bartles categorization is one of the best known and widely used, it has
been subtracted from a discussion among the wizzes6 of MUD2 rather than grounded in
empirical research and therefore might be flawed. In fact, there has been a lot of criticism on
his typology. One of the main issues raised is the fact that the actions, strategies and goals
that Bartle ascribes to the player types, result from discussion rather than actual behaviour.
For example, Bartle clusters role-playing behaviour with the socialiser player type, but talking
with other players might not even be related to role-playing. Another issue is about the
generalizability of his summary, since it was a discussion by users of one specific MUD only
and one that allowed player killing as well. It might well be that the type of MUD can account
for some playing styles, instead of the specific game content itself, yet Bartle (1996) claims
that his initial player types can be used among all the different types of MUD. All in all there
are several questions that can be posed around Bartles typology.
Wizzes is a gender non-specific term, meaning wizards and witches (Bartle, 2003, p.165) and are the more experienced
players of MUD2.
The Big Five model, the result from several factor analyses on descriptive vocabulary,
became the most popular model for describing personality during the 1980s (McAdams,
1995). The Big Five theory maintains that the personality of people can be described by
rating them on five different dimensions: extraversion, agreeableness, conscientiousness,
emotional stability and openness to experience. All five dimensions are bipolar, meaning that
they are a continuum between two opposites, leaving ten personality types. Costa and
McCrae (1992 in: Howard & Howard, 1995) named the ten resulting types and those can be
found in table 1.1 together with some associated traits taken from the mini-marker test of
Saucier (1994). Interesting to note about table 1.1 is the explorer, which is also in Bartles
typology. This suggests that there might indeed be a link between offline personality traits
and online behaviour.
Table 1.1: Costa and McCraes (1992, in: Howard & Howard, 1995) labels for the opposites on the Big
Five factors and some associated traits taken from Saucier (1994).
Positive (traits)
Negative (traits)
Extraversion
Agreeableness
Conscientiousness
Emotional Stability
Openness to Experience
Are there different online playing styles of MUD players in general and can
these playing styles be related to offline personality traits?
This research question can be divided into two parts: the online playing styles of MUD
players in general with Bartles (1996;2003) typology and Yees (2006a) research on player
motivation as a starting point, and the offline personality traits of the players as measured by
the Big Five mini marker test of Saucier (1994) in combination with other explaining variables
as demographics and MUD and player characteristics. To investigate the relationship
between these two parts this thesis uses two sub-questions:
How are these online playing styles related to the offline personality
traits of the players and other explaining variables as MUD and player
characteristics?
The main goal of this thesis is to find a categorization for different types of playing
styles and to relate online behaviour to offline personality traits. If Bartles initial four player
types can be reproduced, it will strengthen his theorising on them. However, if this research
leads to different outcomes, this thesis will try to revise the theory to match the findings. Also
the combination between online behaviour and offline personality traits will lead to a more
scientific grounding of the player types literature.
The scientific relevance of this thesis will be to add to the literature on player types
and interests of MUD players. As indicated before, the existing model is based upon
theorizing from a discussion, but only partly tested empirically. The research will try to
provide a conclusion to the theory. The link between online and offline traits could give
insight in the build-up of the playing styles and provide interesting knowledge for further
research. The practical relevance will be to provide some insights into a rapidly growing
phenomenon and to present a working model of online player types that can be used as
indication for target group marketing and management. Subsequently, the link between
online playing styles and offline, Big Five, personality traits, like agreeableness and
emotional stability, might give insight in which areas to invest to attract players.
The structure of the thesis is as follows: chapter two will discuss the theoretical
framework about player types and interests and deals with the offline personality traits of the
Big Five, leading to the conceptual model. Chapter three will discuss the research method:
an online survey followed by an exploratory principal component analysis on the player
interests to find underlying player types and subsequent multiple regression analyses to
relate the found player types to Big Five offline character traits and the other variables.
Chapter four will contain the results of the research and a thorough discussion of the
findings. Chapter five will provide a conclusion which will link back to the research question
posed above and will contain recommendations for further research.
According to Bartle (2003), there are two reasons why it is important to know who plays your
MUD and why they are playing it, both being marketing issues. Firstly, because if you know
which people like to play your MUD, you could search for newbies that match your MUD and
secondly, if you know which type of players you want, you can revise your MUD in order to
make it appeal to your targeted player base. Demographics are useful statistics, but mostly
demographics mean generalizations, generalizations mean stereotypes and stereotypes
mean problems (Bartle, 2003, p.126). As stated in the introduction, people play MUDs to
have fun, so what people do in the MUD depends on which activities they enjoy. Therefore it
would be much more useful to have a model that categorizes players on what they do in the
MUD, rather than who they are outside the MUD. Bartle (1996; 2003) provides such a model,
although it is not rooted in science. His original model is the subject of this research.
The following paragraphs hold the theory for this thesis. The first paragraph will deal
with the initial typology of Bartle (1996), classifying players into four categories: achiever,
killer, socialiser and explorer. The core of the theoretical body consists of these player types.
Also this paragraph will deal shortly with the Bartle test: a test that supposedly measures
which type of player you are. The second deals with criticism on Bartles (1996; 2003)
typology. The main rival for Bartles player types is the research of Yee (2006a) about the
motivations of MMORPG players. Although graphical MMORPGs are not exactly the same
as text-based MUDs, Yee provides a different insight by focussing on motivations. His factor
analysis does not reproduce Bartles whole model and this might give an interesting twist to
the theory. One of the main issues about Bartles model is his claim that it can be used for
any kind of MUD. A fair amount of MUD users and other researchers argue that it only goes
for the game-like MUDs and that players of social MUDs, for example, will always fall into the
socialiser type and that role-playing is hardly mentioned.
The remainder of the chapter will deal with alternative ways of classifying players and
motivations for playing (paragraph 2.3) and will deal with the Big Five model as a
measurement for offline personality traits and possible links with online behaviour in MUDs
(paragraph 2.4). Lazzaro (2004) as well as Fullerton et al. (2004) report different playing
styles and motivations. Their theories will be discussed briefly as well as a few hierarchical
classifications of players like Farmers (1992) and Hedrons (1998). The theory behind the
Big Five model will be explained briefly and after that the separate factors will be discussed
and related to certain behaviours and characteristics of Bartles player types. The final
paragraph is reserved for the conceptual model and the hypotheses. This research is
foremost an exploratory research, but with an indicated direction based upon Bartles theory.
Therefore it is hard to formulate specific hypotheses. The conceptual model gives some
indication about how some things are related, but other than that it would have no added
value to define specific hypotheses. This will be explained further in paragraph 2.5.
Bartle (2003) agrees with the fact that the word killer is rather unfortunate. Griefer may be a better term, following Foo and
Koivisto (2004)
be won since it is a persistent world, so they seek their fun in being the best. The exact
behaviour is dependent on the kind of game. If gaining levels and becoming powerful are the
main goals in the game, the achiever will try to kill everything that moves to gain experience
and equipment. If the main goal is player killing, an achiever may set out to hunt every player
to score points. And if the goal is enforced role-playing, the achiever will make sure he is
never out-of-character. Achievers are always busy with the outcome of their actions and if
those actions help them to reach their goals. According to Bartle (1996, p.3) they say things
like: Sure Ill help you. What do I get? and Only 4211 point to go. In essence, achievers
are the only ones that really see MUDs as games (Bartle 2003).
The second player type is the explorer. Explorers try to find out as much as they can
about the virtual world (Bartle, 1996, p.3). The main interest for the explorer is the game
content, which will lead them to start out exploring the breadth of the MUD (the areas and
maps) before moving on towards puzzling out the depth of the MUD (the game mechanics).
Their behaviour is aimed at the discovery of new things, areas, workings and even bugs
through which they understand the world of the game more fully. A typical explorer thing to
say would be: Hmm or Why is it that if you carry the uranium you get radiation sickness,
and if you put it in a bag you still get it, but if you put it in a bag and drop it then wait 20
seconds and pick it up again you dont? (Bartle, 1996, p.4). Explorers see the game as a
pastime, like reading (Bartle, 2003).
The third player type that Bartle (1996) distinguishes is the socialiser. Socialisers use
the games communicative facilities, and apply the role-playing that these engender, as a
context in which to converse (and otherwise interact) with their fellow players (p.3).
Socialisers are concerned with people, not just the players in game. They like to interact with
the people behind the character by listening, sharing information and emphasizing. Their
main interest is to know people and form worthwhile relationships with them. Phrases
associated with socialisers according to Bartle (1996, p.4) are: Hi! or Yeah, well, Im having
trouble with my boyfriend.. The MUD is entertainment to these players (Bartle 2003).
The final player type mentioned by Bartle (1996) is the killer. Killers use the tools
provided by the game to cause distress to (or, in rare circumstances, to help) other players
(p.3). These kinds of players like to impose themselves upon others, mostly causing distress
in their wake. In some cases it might move towards helping, although the good feeling
resulting from it is insubstantial to the feelings of power and influence by dominating other
players. This behaviour often results in killing other characters, which gives it the name, but
the underlying motive is the joy of doing it. According to Bartle (1996, p. 5), killers are people
of few words by saying things like: Coward! and Die. Killers see the MUD as sport, like
hunting and fishing.
ACTING
|
|
|
killers
|
achievers
|
|
|
|
PLAYERS ------------------+------------------- WORLD
|
|
|
|
socialisers
|
explorers
|
|
|
INTERACTING
After having put the model together, Bartle theorized about dynamics between the
four types, focussing on the effect of increasing or decreasing the number of any single type.
For example increasing the amount of socialisers will lead to more killers and more
socialisers and a decreasing amount of achievers leads to fewer killers and a slight decrease
in socialisers. Although these dynamics described by Bartle might be very useful, they are
rather irrelevant for this research, because this research is about validating the grounds on
which the dynamics are based. If the player types can be validated, it will lend strength to the
different dynamics. But if the player types are different or can not be found, the dynamics
might need serious revision or can even be put aside. An important part that Bartle mentions
10
with the dynamics though is the balance between the player types in a MUD. If a MUD
succeeds to attract a myriad of the different player types, it provides the best way of survival
and success for the MUD.
2.1.3 The Bartle Test
In response to Bartles (1996) article about player types, Andreasen and Downey wrote the
Bartle test (mentioned in Bartle, 2003), which was about testing the theory. Players can take
an online, thirty item, binary-choice questionnaire8, resulting in a percentage per player type.
For example, a player that scores 33% on Achiever, 47% on Explorer, 47% on Killer and
73% on Socialiser will be denoted as SEKA9. As Bartle has nothing to do with the
development of the test, he himself looks critical towards it. Some critique on the Bartle test
mentioned in Bartle (2003) includes the self-selecting nature of the participants, the lack of a
neither answer, answer options that apply to two or more player types and the fact that ties
are not handled very well (see the above example: It could as well be SKEA instead of
SEKA, although those could be difference player types).
Next to these critical issues, the main point is left out: This test is designed to
measure the theory of Bartle. It has proven to be reliable, giving the same Bartle Quotient to
the players when they are filling out the questionnaire the same way, but measuring your
intelligence by measuring the size of your head is also reliable, but most authors would not
agree that you are measuring intelligence that way. It is the validity of the Bartle test that
might be an issue. Although Bartles (1996) categorization might be a good representation of
player types, there have been several criticisms on his categorization and the grounding in
experience rather than science. This criticism is addressed in the next paragraph.
The Bartle test used to be at www.andreasen.org/bartle, but it was taken over by the Guildcafe and is now hosted on
www.guildcafe.com/zQuest.php.
9
This is the score obtained by selecting the first of the two options on every question.
11
behavioural and motivational items together into factors, trying to replicate Bartles model.
This is much like this thesis will try, although Yees focus was on motivations for playing
instead of playing styles, which is not fully compliant to Bartles model.
Another critique, partly addressed by Bartle himself, is the social versus game-like
debate. Bartle considers his player types valid for all different types of MUD focussing on the
overall behaviour in MUDs. Bartles reasoning is straightforward and has similarities with the
ambiguity of the term MUD as mentioned in the introduction. This also relates to another
issue: are there different player types or are there players of different MUDs instead. This
paragraph will deal with the above mentioned criticisms. First it will look into Yees (2006a)
pioneering work on MMORPGs (2.2.1) and secondly, the social versus game-like debate and
its consequences for the player types will be addressed (2.2.2).
2.2.1 Yees Player Motivations
As stated in the introduction, MUDs are the predecessors of MMORPGs and have a lot in
common with them. Yee (2006a) studied the motivation of MMORPG (or MMOG) users.
However, Yees motivational approach is not the same as Bartles playing styles. A
motivation is the reason why someone does something, behaviour is the actual doing it.
Bartles player types is about what players like to do in a MUD and Yees player motivations
are why someone plays an MMORPG. Nevertheless, Yee used behavioural items in his test
(for example: I usually dont chat much with group members and I beg for money and
items in the game , Yee, 2006, p.46-47) and Bartles typology also involves some
motivational issues since it came from the question: What do people want out of a MUD
(Bartle, 1996, p.2) focusing upon the reward for people. So, although this distinction exists
and needs to be kept in mind, there are several similarities between Yees five motivational
factors and Bartles four player types.
The five10 motivations for playing MMORPGs by Yee (2006a) are summed up in table
2.1, displaying the corresponding items loading on these factors.
Yees five factor model is the best known and therefore the one used in this thesis. Different articles by Yee, however,
mention differing numbers of factors. For instance Yee (2004) lists six factors, breaking up relationships in a casual and
serious socialiser, and Yee (2006b) reports a ten factor model that can be clustered into three groups: Achievement, Social
and Immersion. These differences are probably due to the series of online surveys that Yee conducted.
12
achievers. The only real distinctions between Bartle and Yee are the explorer player type and
the immersion motivation.
Table 2.1: Yees Player Motivations and related test items (based upon Yee, 2006, p.47-48)
Motivation
Relationship
Manipulation
Immersion
Escapism
Achievement
As mentioned before, one of the criticisms on Bartle is that he does not deal with roleplaying. The immersion factor in this motivational model, might indicate that this is indeed a
deficit in Bartles typology. Bartle mentions role-playing somewhat in the socialiser type, but
Yees results seem to indicate a more independent motivation. The fact that people try out
new roles and want to feel themselves a part of the story is not addressed by Bartle. Bartle
(1996) notes, however, that he favours the view that role-playing is merely a strong
framework within which the four types of player still operate (p.22). With Yees immersion
motivation in mind and the load of encouraged and enforced role-playing MUDs available14,
this study will attempt to see if a separate role-player type can be found or if Bartles
argument holds. This is also closely related to type of MUD and the social versus game-like
debate (see below).
On top of this new immersion motivation, the lack of discovering the explorer
motivation is also an interesting case, since it is one of the archetypes of Bartle (1996).
Several explanations seem plausible. First of all, an earlier analysis (Yee, 2004) lead to the
11
RL stands for real life. Yee (2006a) seems to assume that everyone is familiar with this by putting it in an item.
Negatively correlated to the immersion motivation.
13
XP stands for experience, or experience points. Again Yee (2006a) seems to assume familiarity with the term.
14
The Mud Connector lists over 300 enforced role-playing MUDs and 650+ encouraged role-playing MUDs. (figures July 2007)
12
13
conclusion that the explorer, although several items should have measured him/her, was not
important enough on its own account. However, some of the items intended for the explorer
did correlate some with achievement. This might indicate that exploring in itself is not a
playing style, but rather a means to achievement.
A second explanation for the lack of the explorer might be that the knowledge of
game mechanics and workings is only highly interesting to game designers and developers.
Only immortals, coders or builders might be interested in game mechanics. In MUDs, it is
rather common to advance up to a higher standard. Several MUD administrators and coders
have at some time played their own or other MUDs (see also the hierarchical career paths
mentioned in the next paragraph). It might well be that the ones that want to design and
develop or are interested in workings and game mechanics are the ones that are more likely
to advance. For MMOGs this works slightly different, because they are run by companies
and have a larger player base. Although there might be some advancement in MMORPGs,
they remain a paid service, and the company pays their builders, developers and coders to
create content. These people are hired and are not always raised from playing the game
itself (next to the fact that there are far more players in the game competing to become
content creators). This is exactly the reason why explorers might be found on MUDs, but lack
in MMOGs.
Although the above is highly hypothetical, it needs to be taken into account when
MUDs are considered. Therefore this thesis uses Bartles MUD player types as the base rate
and Yees MMORPG motivations as additional information. The exact methodical work-out
can be found in chapter three.
2.2.2 The Social Versus Game-Like Debate
According to Bartle (1996), the development of the TinyMUD code, which completely lacked
a combat model, lead to a small schism between social MUDs and game-like MUDs. This
has, as displayed in the introduction, lead to different names for different kind of MUDs, with
the term MUD related to the adventure game types and terms like MOO and MUSH related
to socially oriented MUDs. Bartle argues, however, that although the social MUDs might be a
different branch of the MUD tree, and a major one too, they are nevertheless part of it
anyway (See Keegan, 2003). This thesis already supported using the term MUD as an
overarching concept, so in this it agrees with Bartle.
However, Bartle (1996) also argues that the player types are useful for social MUDs
and game-like MUDs alike. He focuses on the fact that social MUDs do have their killers,
although it manifests itself more griefer-like, due to the lack of options to do it in a head-on
fight. He mentions several negative occurrences of grief play, famous in MUD research
history, like the virtual rape mentioned by Dibbell (1998) and the deliberate, terrorism like,
14
fracturing of a community by Whitlock (n.d.). Also Bartle mentions that social MUDs have
their achievers, for instance people that want to build the best area or want to possess the
best rooms on the MUD. This seems a valid point, but the MUD community does not fall for
this it seems.
Also game-like is a very broad range and contains two subcategories. Role-playing
games are games that are driven by playing a specific role in the game. The players are
supposed to stay close to the story of the game and the story behind their online characters.
Player-killing games are games that allow the online killing of other real life players instead
of only killing computer operated monsters in the game. These different games are not as
separable as they look in this respect, because MUDs combine different levels of both. A
MUD could offer restricted player-killing in an enforced role-playing environment. This could,
for instance, involve a player playing a vampire, that can attack other players, but only if
he/she is hungry in the game. Other MUDs allow pure player killing and only accept roleplaying behaviour if people feel like it. The main goal of the MUD becomes killing others in
that respect.
As posed in this way, it is not surprising that this might have an effect on the player
types. An achiever on a pure player-killing MUD would be an ideal killer type and
achievement within an enforced role-playing MUD would be to stay as close to the character,
minimizing out-of-character behaviour. Therefore it will be interesting to see if the player
types are related to the different MUD types or if the claim of Bartle holds that the different
MUDs will also find their specific player type.
15
15
16
17
The hierarchical classifications are relevant in another way. Bartle (2003) proposes a
relatively standard career path through the player types (killer -> explorer -> achiever ->
socialiser)16. The hierarchical classifications do not fully agree with Bartle though. Farmers
(1992) model fits most to Bartles career path. Passives are the newbies and hardly involved
players. Actives are the killers, explorers and achievers. Motivators and caretakers are
socialisers. Geek gods fall outside Bartles model. Hedrons (1998) model has a different
path. He also starts out with newbies in the survival and competence circles (in which
competence could have a bit of explorer inside). Then they advance to beating the game.
Comparing this to Bartle this would be the achiever. Proving mastery could be done in two
ways according to Hedron: either helping other players (this should be a more socialiser
approach) or killing every threat (as a killer would). After this stage a time of exploration
starts and all kinds of possible ways to have fun in the game will be attempted. This is the
exact explorer as Bartle sees it, therefore Bartles career path should be different: achiever ->
socialiser or killer -> explorer. To check for at least some development in the game the years
that someone plays a MUD will be taken into account.
16
Bartle does not include newbies in his proposed career path. Newbies should first learn the game somewhat and only then
they will start falling into one of the four player types.
18
The Big Five emerged during the 1980s, when the heavy preliminary work of Allport
and Odbert (1936 as mentioned in McAdams, 1995) finally got recognized for its value.
Allport and his colleagues used a lexical approach and catalogued almost 18,000 terms
from a 550,000 word dictionary that referred to psychological traits and evaluations. Catell
(1943 as mentioned in Goldberg 1990) started to construct 171, mostly bipolar, scales, using
the terms of Allport. By letting people use the scales to describe others, he reduced the
number of terms to about 35 bipolar clusters. Several subsequent research and replications
by various researchers (See for example Goldberg, 1990, or McAdams, 1995 for an overview
of names) revealed that only five factors could be replicated over all.
During the behaviourist days in the 1960s and 1970s the Big Five never found true
acceptance, because the behaviourists focus was on stimulus and response instead of
traits. With the rise of cognitive science at the start of the 1980s that view dwindled and traits
came back to the fore. During the 1980s and 1990s, there was extensive research on the
five factor model, amongst them Goldberg (1990) and Costa and McCrae (1992 as
mentioned in Howard & Howard, 1995). The five factors were also found in different
languages thereafter (for example in Dutch, German and even Chinese and Japanese
language, all mentioned in McAdams, 1995), giving strength to the lexical trait taxonomy.
All five factors can be seen as clusters of trait descriptions, like five buckets filled with
corresponding descriptive words and terms. Although the five factors themselves have been
assigned different names, indicating disagreement over the meaning of certain factors, the
overall picture (or the content of the buckets so to speak) is nevertheless surprisingly the
same. Because this research will use the mini-marker test of Saucier (1994), this thesis will
use his assigned names of the Big Five factors, which are heavily based upon Goldbergs
(1990) labels. I: Extraversion (2.4.1), II: Agreeableness (2.4.2), III: Conscientiousness (2.4.3),
IV: Emotional Stability (2.4.4) and V: Intellect, Openness, or Imagination (2.4.5). All five will
be discussed shortly in the remainder of the paragraph along with possible links with the
player types.
2.4.1 Extraversion
The extraversion factor, sometimes called surgency, encompasses items like spirit,
expressiveness and spontaneity (Goldberg, 1990). Extraversion is seen as one end of the
spectrum of this factor with introversion on the opposite side. Items related to introversion
are for example aloofness, silence and shyness. (Goldberg, 1990). People scoring high on
extraversion talk more often and sooner, and tend to make more eye-contact with other
people than introverts. People scoring low on extraversion (introverts) on the other hand are
more solitary and are more attracted to artists, mathematicians and engineers (McAdams,
1995).
19
McAdams (1995) reports several studies that indicate important and interesting
conclusions about the extraversion factor. First of all, it is relatively stable. Conley (1985b: in
McAdams, 1995) came to this conclusion after a fifty year longitudinal study, which showed
extreme consistency in extraversion. Secondary, extraverts report more positive feelings.
This does not mean that they report less negative feelings, however. Part of this might be
explained by the fact that extraverts respond less to punishments than introverts. A study by
Pearce-McCall and Newman (1986: in McAdams, 1995) revealed that extraverts even seem
to taunt the punishment by continuing in a more risky way after experiencing punishment.
The third conclusion about the extraversion factor is that extraverts seem to have a lower
state of arousal given standard conditions, which lead Eysenck (1967: in McAdams, 1995) to
believe that extraversion/introversion is rooted in human nature. Both extraverts and
introverts strive for an optimal amount of arousal, but due to a higher arousal standard in
introverts, they are unable to tolerate the same increase in arousal that extraverts can. This
might also explain why introverts will try to avoid stimulating environments and extraverts on
the contrary tend to seek them out.
It might well be that extraverts tend towards the player side of Bartles interest graph,
because they are good in the interaction between human beings, and introverts towards the
world side, since they are more solitary and calculating. Extraverts might risk themselves
after a nasty experience (showing killer behaviour), and introverts might minimize their
additional arousal, by avoiding humans and start out to explore the world, on their own pace
(showing explorer behaviour).
2.4.2 Agreeableness
The agreeableness factor harbours items like cooperation, amiability and empathy, opposing
the other end of the factor with items as overcriticalness, bossiness and rudeness (Goldberg,
1990). According to Howard and Howard (1995) people that score high on agreeableness
(adapters) are people that will march to the drumbeat of many different drummers, while low
agreeableness persons march only to their own drumbeat (p.5). McAdams (1995) notes a
low agreeableness (challenger) is the worst thing that can happen to you by words: they are
malicious and untrustworthy and do not regard for the feelings of other people. Along with
conscientiousness, there has been less research on this factor than on the other three.
Overall, this factor seems to be linked to the need for affiliation and need for nurturance
scales (McAdams, 1995) and is generally matched with love.
Agreeableness, along with extraversion, seems to relate to the player versus world
axis of Bartles model with the adapters at the player end and the challengers on the world
end. Looking more closely, the action-interaction axis might be a better predictor with killers
20
and achievers being less agreeable and socialisers and role-players/explorers as more
agreeable (towards the players or the world respectively).
2.4.3 Conscientiousness
The conscientious factor is positively linked with items like organization, efficiency and
precision, while negative terms associated with conscientiousness are disorganization,
inconsistently and forgetfulness (Goldberg, 1990). People scoring high on conscientiousness
are labelled focussed and those scoring low flexible (Howard & Howard, 1995). While these
terms both seem at least to some extent positive, the negative traits associated with flexible
terms them highly unreliable people. Focussed persons are goal-oriented and work towards
concise overviews and never loose the overall picture. So this makes them reliable. As
mentioned at the agreeableness factor, there has also been less empirical research on this
factor (McAdams, 1995), but it is often matched with work or work-related issues.
Combining this factor with the player types of Bartle, every style of play seems to
work to some related goal: socialisers want many relations, explorers/role-players like to
submit to or interact with the world, achievers want to become the best and killers want to
create havoc and mayhem. Although all types have their subsequent goals, the one most
related to the game itself will be the achiever, which makes achievers a possible match to the
conscientious factor, although role-players in an enforced role-playing environment or killers
on a pure player-killing MUD could be seen as such as well.
2.4.4 Emotional Stability
This factor has often been called neuroticism or negative emotionality, although that is
inconsistent with the other factors, placing the emphasis on the negative part of the factor.
This might well be because Goldbergs (1990) factor analysis (although he labels this factor
positively as emotional stability) only shows two positive items: placidity and independence.
Some of the negative items related to this factor are insecurity, instability and emotionality
(Goldberg, 1990). The labels put to people scoring high or low on emotional stability are
resilient and reactive, with resilient people being more rational about the world around them,
sometimes even impervious17, and reactive people being more unhappy with their life and
experiencing more negative emotions than most people (Howard & Howard, 1995).
Part of the conclusions on extraversion can be found as well in the emotional stability
factor (McAdams, 1995). Emotional stability is also stable over time, which was concluded in
the same fifty year longitudinal study (Conley, 1985: in McAdams, 1995). Also people that
report a low emotional stability report a higher amount of negative feelings, which again does
17
Howard and Howard (1995) supply a nice anecdote about a non-reactionist response: Their choir director did not miss a beat
during a dress rehearsal when the podium on which he was standing collapsed forward. He just placed his feet at angles
like a snow plow and kept his baton moving (p.4), while all instrumentalists and singers laughed their head off.
21
not indicate that these people report less positive feelings. So while extraverts experience
more positive feelings while there seems no effect on negative feelings, it is the opposite for
reactives. Eysenck (1967: in McAdams, 1995) also theorized that emotional stability has to
do with human nature, but although there is some support (For instance Davidson &
Tomarken, 1989: in McAdams, 1995), the research is not consistent on the findings.
Emotional Instability is linked with a lot of negative and bad feelings. Research by
Bolger and Schilling (1991: in McAdams, 1995) provides some insight in the reasons. First of
all people with a low emotional stability reported more daily stressors. This was not due to
oversensitivity for negative affects, because their spouses supported the accuracy of the
reports. Low scores on stability seem to expose individuals to more stressful events. A
second conclusion was that emotionally unstable persons not only reported more stressors,
but that they also reacted more strongly upon them. Bolger and Schilling argue that this is
due to ineffective coping strategies: highly neurotic people are less able to take the matter
into their own hand and fall victim to their anxiety and depression. The last conclusion was
that the negative feelings of emotionally unstable persons can not be explained away by their
daily experiences. Overall it can be said that even when the day does not bring bad
experiences, neuroticism provides bad feelings.
Relating the above factor to the player types, there seems hardly any overlap on any
of the player styles. The only qualifiers might be the victims of killers, which could be any
type of player. On the whole, negative experiences will occur though (see for instance the
part about what constitutes fun for players, mentioned in the introduction) and it will probably
depend on the player him- or herself whether they can cope with them, rather than their
playing style.
22
imagination, called preservers (Howard & Howard, 1995), have narrower interests and more
conventional views. Put simply, people that score low on this factor are more worldly and
traditional than people that score higher.
Putting this in relation to MUDs and the player types by Bartle, it must be noted that
this seems a highly relevant factor. Because of the text-based nature of MUDs fantasy and
imagination are needed to see through the text. Also an open mind, being able to distinguish
that mere text can provide a worthwhile gaming experience, seems mandatory. It would
therefore probably be no surprise if most players score on the positive side of this trait to
begin with. Nevertheless, this factor will probably link more to the world side of Bartles model
and still even more with the action/interaction axis: so the explorer and/or role-player. As
McAdams (1995, p.288) puts it: absorption is associated with an intense and vivid fantasy
live. Persons who score high on absorption may become so immersed in their imaginative
experiences that they lose all track of time, place and identity. This links closely to
immersion in the game environment.
paragraph 2.3 did not add much with Lazzaro (2004) naming four motives, pretty congruent
to Bartle and Yee and Fullerton et al (2004) just mentioning possible types that again have a
lot in common with the types mentioned by Bartle and Yee. The hierarchical classifications,
although they seem rather irrelevant for the purposes of this thesis, could provide interesting
findings if combined with Bartles proposed career path.
With so many overlap between the player styles and motivations, this research will
look for the four original types proposed by Bartle. Only one additional player type based
upon Yees immersion motivation (see paragraph 2.2.1), which differs greatly from the
concept of the socialiser as Bartle proposed, will be added to see if this holds. Combining the
immersion factor with Bartles four quadrant model, it might be interesting to see if the roleplayer fits the spot of the explorer if the explorer does not show up, since role-playing
involves interaction with the story/world. Although this might be interesting to look at later the
main goal will be to try to reproduce these five player types: achiever, explorer, role-player,
socialiser and killer/griefer.
After the player types, the theory started looking into the offline personality traits of
the Big Five. The Big Five factors (extraversion, agreeableness, conscientiousness,
emotional stability and intellect, openness, or imagination) were shortly described and a
possible link between the factors and the player types of Bartle was suggested. After
determining if and which player types can be abstracted from the data, it will be interesting to
see if those player types are linked to offline personality. The Big Five will, next to more
standard variables like demographics and MUD and player characteristics, be used to draw
conclusions about the player types. If the player types can be reproduced, there might be
interesting links between online and offline. It would be useful to see if the offline personality
can be used to predict online behaviour.
24
instance. And a MUD set up for educational purposes will probably attract a very different
player base. Also people that have played for an extensive amount of years will develop
another view on a MUD than players who have just started to discover MUDs and players
that play multiple MUDs might have a different playing style on either one. Of course, next to
the above explaining variables, part will probably be explained by more standard
independent demographic variables like sex, age, education and marital status.
Figure 2.3: Rough conceptual model
As mentioned before, it is hard to pose specific and very clear hypotheses in this
exploratory research. Nevertheless there are a few expectations that can be mentioned
beforehand. Of course it is expected to be able to find at least some of the mentioned player
types. The achiever, the socialiser and the killer/griefer show to be remarkably stable over
the categorizations. The explorer and the role-player are also expected to be found, although
they might differ a bit from Bartles explorer or Yees immersion motivation. Next to this, it is
expected that the independent variables of MUD and player characteristics together with
demographic variables will explain at least a substantial part of the difference in playing style.
The link between offline personality and online player type distinction is expected to explain
an additional part of the differences in playing style.
25
3. Research Method
The player types were never studied thoroughly and are hardly grounded in empirical data.
To enable a firm empirical analysis it was necessary to get as many respondents as
possible. Also it was deemed better to try to reproduce the player types first than to attempt
to colour the picture based upon the current player types. A qualitative approach would not
be sufficient because of its limitations for sample size and the extensive data that would have
to be analysed in the given time frame. A quantitative approach would be much more
convenient, as well for the possible number of MUDs and different players that could be
studied as well as for the ease of analysing. Furthermore, it could allow the number of
respondents for a principal component analysis, much like the method Yee used to find his
player motivations.
Therefore this thesis chose to use an online survey as the quantitative research
method. Several reasons can be mentioned for making it an online survey. The first and most
important one is the fact that the targeted research subjects are people playing an online
game. Internet access would be no problem, because people need to have it anyway in order
to be able to play a MUD. Secondly, it is supposed to be a global study and sending letters
all over the world would not only be very expensive, but there were no specific home
addresses available as well. Additional reasons were the ease of and small amount of time
needed for filling it out and the possibility of a large number of respondents.
26
Five items (part C, 3.1.3). After filling out the last page with Big Five factors, people had the
opportunity to submit their email-address to receive a copy or a summary of the research
and/or if they wanted a chance on a gift certificate of $15 for www.amazon.com. Three lots
were drawn from all the respondents that had answered the survey seriously and fully and
indicated to be interested in the gift certificate reward. The overall time associated with the
survey was about ten minutes and a counter at the top of the page showed how many pages
were still to come to let the participants know their progress. Nevertheless, several
respondents did not finish the complete survey. If there was enough information submitted,
the respondent was kept inside the sample and the rest was labelled as missing values.
Appendix A shows the complete survey as it was hosted on the server, complete with the
introductory text and thank-you-pages.
3.1.1 Part A: General Information
The first part of the survey, about general information, consisted of fifteen questions, spread
out over four pages. The first two pages contained questions about demographic variables.
The first contained four questions: gender (closed: male/female), age in years (open: max 3
digits), country of origin (closed: drop-down-menu with 255 different countries) and
occupational
status
(closed:
full-time/part-time/student/homemaker/unemployed/retired).
Both gender and occupational status had HTML coded radio buttons that allowed only one
answer and the drop-down-menu of country of origin could only select one of options (the
default option was: Please select your country). The second page contained three
questions: education (closed: primary/secondary/higher/none of the above), marital status
(closed: single/married/divorced) and if the respondent had children (closed: no/yes). All
three questions were with HTML coded radio buttons, allowing only one answer. Due to the
difficulty of education in a global survey it was kept simple by only dividing it in three options
and a none option. Higher education also contained university. The marital status had a
specific divorced option, although such a person would also be single. The answer options
also listed not living together and living together behind single and married to cover all
possibilities.
The third page was mainly about player characteristics and contained four questions:
how many hours a week the respondent played (open: max 3 digits), if the respondent
played multiple MUDs (closed: no/yes), for how many years the respondent had played
MUDs (open: max 2 digits) and what the usual business was on the MUD (closed, more
options possible: player/builder/wizard/administrator/coder). The multiple MUDs question had
radio buttons of which only one could be selected and the usual business question had five
checkboxes which allowed multiple answers. The answer options for the usual business on
27
MUDs was taken from the MUD players directory signup (http://www.mudconnect.com/mpd/),
although the order was altered.
The fourth page also contained four questions, but those were about MUD
characteristics: The full name of the respondents main MUD (open: max 60 characters), if
the respondents main MUD offered role-playing (closed: no RP/accepted RP/encouraged
RP/enforced RP), if the respondents main MUD offered player-killing (closed: no
PK/restricted PK/plenty PK/pure PK) and if the respondents main MUD was educational or
research-oriented (closed: no/yes). The last three had radio buttons as answer options to
allow only one answer. The main MUD was necessary to be able to evaluate if the sample
contained players of different MUDs or if they came from one MUD in particular. The three
questions about the kind of MUD needed to be posed apart from each other, because MUDs
combine different levels of PK and RP.
The different levels of RP and PK were taken from the MUD connector
(www.mudconnect.com) that rates all their MUDs on these PK and RP levels. A MUD with
restricted player-killing is allowing PK only in specific instances, for instance in specific
events or only if both players choose to be player killers. Plenty of PK and pure PK MUDs
are MUDs that do not put a lot of restrictions on the player killing and player killing might
always be a threat to any player. For role-playing, accepted role-playing MUDs are MUDs
that put the stress on freedom to express role-playing behaviour if people like that.
Encouraged and enforced role-playing MUDs put more focus on the story and role-playing
components, with often rewards for staying in character and punishments for out-of-character
actions. Next to a rating on player-killing and role-playing the MUD connector also mentions
if a MUD is educational or research oriented. A lot of these latter MUDs are social in nature,
for instance teaching how to code or how to build areas for MUDs.
The above questions all served as background information and were intended to
measure the smaller squares of the conceptual model. Most of the questions did what they
intended, only the specific MUD characteristic questions about role-playing, player-killing and
educational MUDs gave very different answers per MUD. The same MUD, for example, was
rated as enforced role-playing and the opposite of no role-playing by different respondents.
Probably this was due to the experience of players. This information might be valuable, but
did not serve the purpose of the study. Therefore, to create a reliable measure per MUD for
these characteristics, the submitted MUD names were matched with the entries that exist for
all the different MUDs on www.mudconnect.com and www.mudmagic.com. These entries are
submitted by the MUD administrators themselves and therefore closely match the
administrators view of the MUD. With this method it was also determined if a MUD was
socially oriented or not. With the general information settled the two biggest parts of the
conceptual model needed to be measured.
28
Although this reply might be valid in the case of this particular MUD, there were roleplaying items in the survey and those might have been related to this particular MUD. Part of
these problems relate to multi-stage sampling, which will be dealt with in paragraph 3.2 about
the sampling method.
18
Of course these replies were again answered by the author to urge them to reconsider and participate to enable drawing valid
conclusions about it, whether Bartle sees it wrongly or not.
29
Positive Items
Talkative, Extraverted,
Bold, Energetic
Sympathetic, Warm,
Kind, Cooperative
Organized, Efficient, Systematic,
Practical
Unenvious, Relaxed
Creative, Imaginative, Philosophical,
Intellectual, Complex, Deep
Negative Items
Quiet, Bashful,
Withdrawn
Cold, Unsympathetic,
Rude, Harsh
Disorganized, Sloppy, Inefficient,
Careless
Moody, Jealous, Temperamental,
Envious, Touchy, Fretful
Uncreative, Unintellectual
asking for their cooperation in spreading the message about the survey on their own MUD.
Over 1600 emails were sent separately, correcting for double email-addresses as much as
possible19. In a way this was a multi-level model of sampling.
This multi-level sampling provides its own problems. The main ones being nonresponse and selection at the gate. The first problem was that almost a quarter of the sent
emails were returned because the email-address (or MUD) did not exist any longer. A few
did reply, but mentioned that the MUD that was targeted had died years ago. Selection at the
gate was even worse, leading to a different kind of non-response. Some of the administrators
replied with positive messages but a greater part of them did not think that their MUD should
be in the sample. Mentioning Bartle in the introduction might have been a bad idea, because
administrators often replied that they were familiar with Bartles typology and that it did not
suit their MUD. This is probably due to the social versus game-like debate. Of course all
administrators that replied in this way were urged to still put it through to their player base,
since it would allow statistical control over those types of MUD. The earlier mentioned email
in 3.1.2 about the playing style items not being valid for that particular MUD is one example
of selection at the gate. A reply from another MUD administrator indicates clearly the
problems that some people see in the usage of Bartles terminology:
I don't think this stuff really applies to a MUSH. I've read Richard Bartles stuff about
MUD playing styles and half of them just don't really apply. Anybody who plays long
term on a MUSH is going to be a socialiser in his terminology, the other ones usually
simply aren't applicable (email by an administrator of the Greatest Generation).
Next to the self-selection nature of administrators, the way that they communicate it to
their players could also influence the results. Some administrators were very eager to put it
through to their player base, but others just posted a note in the MUD or forum post on their
website and left it at that. Therefore it is hard to determine if the sampled MUDs and MUD
players are representative of all MUDs and MUD players. To prevent this as well as possible
every MUD entry on www.mudconnect.com was emailed, ensuring that at least every MUD
administrator listed got the email.
The above mentioned steps (posting on the forums and emailing all administrators
listed) lead already to an overwhelming amount of over 1800 respondents. Because time
grew short, both due to the amount of time needed to put it all in the statistical programme
SPSS and the time left in which to analyse the results, there was no follow-up email or forum
post.
19
There were 1673 MUDs listed during the sampling phase (April/May 2007). At the moment, however, maybe due to the
authors mention to the MUD connector about the load of experienced returned unable to deliver emails, there are only
1515 MUDs listed. (figures August 2, 2007)
31
32
(71.0%, N=1736) is working on or has finished a higher education. Nearly twenty four percent
(23.7%) lists a secondary education as their highest completed or current education.
People have been participating on MUDs for an average of 8.97 years (N=1726
SD=3.92) with a range of 0 (probably due to playing less than a year) to 30. Also the age that
people started playing (age - years playing) varies widely with a range of 8 to 58 and an
average age of 19.52 (N=1715, SD=8.02). Almost everyone (94.4%, N=1738) indicates to be
a player occasionally, but almost a third of the sample (31.9%, N=1738) indicates to be more
active on a MUD than being a mere player, for instance as a builder or coder. Mere players
spent a significant (t=-3.193, p=.001) amount of years less (M=8.76, SD=3.94, N=1176) on
MUDs than more active players (M=9.41, SD=3.86, N=549). There is no significant (t=1.566,
p=.118) gender difference between mere players and more active players, but more active
players spent a barely significant (t=-2,078, p=.038) amount of hours more on MUDs
(M=20.24, SD=17.30, N=540) than do mere players (M=18..67, SD=14.28, N=1170).
People, on average, spend an awful lot of hours on their MUD(s): Over nineteen
hours a week (M=19.23, SD= 15.32, N=1710). Moreover, almost fifteen percent (14.8%)
report playing more than thirty hours a week with values as high as ninety and hundred hours
a week. Women significantly (t= -3.833, p<.001) spend more hours on MUDs (M=21,73,
SD=15.31, N=415) than men (M=18.43, SD=15.25, N=1288). Almost seventy two percent
(71.8%, N=1736) are loyal players to one MUD only, but respondents indicating playing
multiple MUDs (M=22.55, SD=16.74, N=485) spend significantly (t=-5.340, p<.001) more
hours in the game than respondents that indicate to play one MUD only (M=17.93,
SD=14.52, N=1223). More than half of the players (56.5%, N=1735) mention being employed
full-time and over a quarter of the sample (26.1%) indicate to be in full-time education. When
occupational status is taken into account, the time spent in the game increases when players
have less time-consuming responsibilities (F=15.900, p<.001) It seems that the more free
time available at home, the more time is spent on playing MUDs. Table 4.1 gives the mean
amount of hours spent on the game, split out by occupational status.
Table 4.1: Average amount of hours spent on a MUD divided for occupational status
Occupational Status
I work full-time
I am a student
I work part-time
I am retired
20
I am a homemaker
I am unemployed
20
N
980
453
144
16
70
72
%
56.5
26.1
8.3
0.9
4.0
4.1
The question in the survey about occupational status did not use I am a homemaker, but I am taking care of my family
and/or home.
33
N
423
289
149
112
57
46
31
30
20
17
%
24.2
16.5
8.5
6.4
3.3
2.6
1.8
1.7
1.1
1.0
Role-Playing
encouraged
encouraged
accepted
accepted
enforced
enforced
accepted
encouraged
encouraged
enforced
Player-Killing
Yes
Yes
Restricted
Restricted
Yes
Yes
Yes
No
Yes
Restricted
To be able to classify the mentioned MUDs, the survey asked for a rating on roleplaying, player-killing and whether the MUD was educational. As mentioned in chapter three,
these figures turned out to be highly unreliable to classify the different MUDs on these
qualities and seem to indicate that these qualities are highly susceptible to the experience of
the player. Therefore the mentioned MUDs were matched with their listed properties on the
MUD connector (www.mudconnect.com). Of these 244 MUDs only seven classify
themselves as educational, and ten could be classified as a social MUD by their theme. This
is probably due to the non-response as mentioned in the sampling method in chapter three.
So unfortunately, these only account for 1.5% (N=1698) of the respondents and it will
therefore be hard to control for those during the player types results (paragraph 4.2). 161 of
the 244 MUDs (66.0%) are listed as an encouraged or enforced role-playing MUD and cover
73.6% (N=1698) of the respondents. 89 out of 244 MUDs (36.5%) classify themselves as
unrestricted or pure player-killing MUD, accounting for 62.3% (N=1698) of the respondents.
21
The fifth (Dark and Shattered Lands) had no entry on their minimum number of players online at all times, but they are the
number one in player votes on the MUD connector at the time of writing (figures august 2, 2007).
34
Of the 244 MUDs mentioned, there are 178 MUDs (73.0%) that have their
geographical location in the United States. Therefore it is not really surprising that the
majority of the respondents comes from the United States (71.2%, N=1741). Second in
number of respondents is Finland (geographical home of BatMUD) with 5.9 percent (N=103).
The respondents come from 49 countries in total and those are closely related to the number
of MUDs that have their geographical location there. A summary of the findings can be found
in table 4.3.
Table 4.3 shows the top nine of the mentioned countries in the survey together with
their associated number and percentage of respondents. Also in the table are the number
and percentage of MUDs that have their geographical location in that country. These two
numbers show a rather similar picture. There are only three countries that seem to diffuse
the ranks and those are Finland (due to the large amount of Finnish players of BatMUD,
which has its geographical location there), Australia, which has only one MUD listed but
seems to have a very scattered player community and Germany, which has five MUDs
listed, but just some players, which might be a language issue, although this remains to be
seen. In correspondence to this, there are several Russian MUDs listed on the MUD
connector (www.mudconnect.com) that indicate Russian as the main language. The relation
between geographical home of the MUD and the nationality of the players is an interesting
topic, moreover because the internet should cross geographical borders and therefore this
might be a nice indication for further research. For this research it is rather irrelevant.
Table 4.3: Division of respondents per country (N=1741) and MUDs per country (N=244)
Rank
1
2
3
5
6
7
8
9
--
Country
United States
Finland
Canada
United Kingdom
Australia
The Netherlands
The Russian Federation
Sweden
Germany
Other
# of resp.
1233
103
85
85
46
25
19
17
14
114
%
70.8
5.9
4.9
4.9
2.6
1.4
1.1
1.0
0.8
6.5
# of MUDs
178
3
14
14
1
7
3
3
5
16
%
73.0
1.2
5.7
5.7
0.0
2.9
1.2
1.2
2.0
6.5
( Rank )
(1)
(6)
(2)
(2)
( 10 )
(4)
(6)
(6)
(3)
(-)
35
in his research (see also chapter three). In the first part of this chapter an exploratory
principal component analysis will be performed to cluster the items again. After the player
types have been defined, the second part will deal with who the people are that fit the player
type by a regression on the sum scores. The distinction between offline and online will be
kept for paragraph 4.3.
36
think that realism in the game is important and I like to try out new roles in the game) and
two intended explorer items (I like to explore all the areas in the game and I like maps,
charts and tables with information about the game). An additional achiever item (I want to
be noted for my achievements) was removed later on, because it loaded slightly on two
components, but did not add to either scale reliability.
The socialiser items provided a different problem. Looking at the correlations only,
four of the items had very low correlations. This would mean only two specific socialiser
items. Looking at the inter-item correlations and alpha they seemed to be a decent scale with
an alpha of .657 and lower alphas when items were deleted. The inter-item correlations were
all above .295, very close to the rule of thumb of .300. For these reasons and to enable a
decent coverage of the socialiser construct they were kept in the component analysis.
The initial principal component analysis was meant to discover the number of
components. Although six components had an Eigen value higher than one, the scree plot
indicated that the sixth factor would not add much to the model and explained just a small
amount of the variance. To be sure, four additional factor analyses were done, two varimax
rotations and two direct oblimin rotations, with either five or six factors. Five components lead
to the best and most interpretable simple structure. The Oblimin rotation (see appendix C for
the complete pattern matrix) produced the best solution, because there appeared to be some
correlation between some components. A disadvantage of using the oblimin rotation in this
occasion is that the component scores for the killer/griefer component are negatively rated,
leading to some kind of anti-griefer. Nevertheless the same items related to each other well
and created reliable scales for the subjectively named achiever type (7 items), role-player
type (7 items), (negative) griefer22 type (6 items), socialiser type (6 items) and explorer type
(also 6 items). An overview of these player types components, their associated traits and
their related reliability can be found in table 4.4.
Table 4.4: The five resulting components, associated traits and their reliability as a scale.
Component
Achiever
Role-Player
Griefer
Socialiser
Explorer
23
Reliability
= .809
= .848
= .818
= .657
= .751
Associated Traits
Power, collect items, rise in levels, rating lists, competition
Role-play, importance of story-(lines), immersion, in-character play
Causing distress, killing, domination, imposition, competition
Meaningful talk, know players, group up, communicate, help players
Find bugs/secrets, know game mechanics, show knowledge
22
The name of griefer instead of killer seems to suit better to the items loading on the component. This is analogous to the term
that Foo & Koivisto (2004) use and even Bartle (2003) agrees with the unfortunate name of killer since it does not always
involve the actual killing of other players.
23
Reliability based on Cronbach's alpha (total item-correlation > 0.295 and alpha if item deleted < scale alpha)
37
As mentioned above, some of the components correlated with the others. The main
correlation is found between the achiever and the explorer: r = .311. This seems analogous
to what Yee (2004) notes in his research about the explorer and achiever being related to
each other, although the explorer is a separate and important component here, contradicting
some of the findings of Yee. This could be, as mentioned in chapter two, be due to the fact
that it is easier to become more than a mere player in MUDs than in MMORPGs.
Two other correlations above .150 can be found. The first between the explorer and
the socialiser (r = .211). This is probably due to the factor loadings of the explorer component
on the "I want to be known for knowledge" and "I like to show my knowledge of the game".
This is of course related to socialising, because it needs others to be recognised and
acknowledged as a knowledgeable person. The second (r = -.160) is found between the
griefer and achiever components. This is probably due to the fact that competition is
favoured by both, but in a different way.
4.2.2 Colouring the Picture: Who are the people behind the components
It is nice to know that the items can be structured in the way as displayed above, but it would
be even better to know which persons fall into these components. To be able to draw
conclusions about this, a multiple regression was done on these components. To provide a
measure for the components, the related items were added up to create a sum score. The
sum score was favoured above component scores for the ease of interpretation24. The
results can be found below. All mentioned regression tables can be found in appendix D to
save up space.
The sum scores for the player types were regressed on thirteen variables, most of
which speak for themselves. A few need further explanation though. The big amount of
higher education and university scores made it more plausible to create a dummy (0=no
higher education, 1=higher education) instead of using the ordinal range. Marital Status was
recoded into a dummy (0=single, 1=together) by adding the people rating divorced as single.
Next
to
these
demographic
variables
the
different
businesses
on
MUDs
Component scores and sum scores, although different scale measures, ought to have a very high correlation and therefore
would produce rather similar results.
38
The multiple regression for the achiever player type can be found in table D.1. For the
demographic variables only gender and age are significant. Men seem to be more of the
achieving type than women and this is also the main predictor for the achiever. This seems
compliant with the somewhat stereotypical view of males as achievers and women as
caretakers. Age also affects the achiever significantly, with younger people scoring higher on
achievement. Having children does have a small positive effect on achievement, but this is
not significant. Marital status and having a higher education are very much insignificant.
Looking at player characteristics, it can be seen that the hours spent playing have a
significant effect. The effect might be the other way around as well though: if you want to
achieve something, you will need to invest time into it. Being more than a mere player is also
significant. The negative relation is not surprising, because advancing is mainly a matter of
the players and not of builders or administrators. The experience on MUDs or playing
multiple MUDs have small effects, but are not significant.
For MUD characteristics it can be seen that playing role-playing and social MUDs
have a rather large negative effect on being an achiever. It seems that these MUDs do not
promote achievement very much. The social MUD variable is not significant though. This is
probably due to the small amount of social MUDs in the sample. Playing player-killing MUDs
is significantly and positively linked with achievement. This is hardly surprising since playerkilling creates competition. Also the insignificance of playing educational MUDs is hardly
surprising since those are targeted at learning and not advancement. Overall gender and age
predict the most of the variance in the achiever type, with the type of MUD (RP and PK the
most) as runner ups. This regression model explains about eleven percent of the variance.
The regression for the role-player can be found in table D.2. Again gender is a
significant and important explaining variable, with more women being a role-player than men.
Age is the second most important of the demographic variables, but has just a small effect
and is not statistically significant. Marital status, having children and being higher educated
are also not significant. These real life demographics might not have much influence on
someone focused on immersion in a MUD. Also the player characteristics predict
insignificant parts of the variation. Only being more that a mere player is positively related to
the role-player. This could be related to the creation of role-playing content (i.e. builders and
administrators), since making up storylines is one of the items.
The main explanation for the role-player type is the players of role-playing MUDs,
which is of course not surprising. If you play an encouraged or enforced role-playing game,
your score on the role-player component goes up by more than six, which is on average
almost a full point on the seven items that are related to role-playing. Educational MUD is
also significant, but negatively related to the role-player. This is obvious, because
39
educational MUDs are about learning things for real life from a virtual environment instead of
immersion in that MUD25. The biggest predictors for the role-player are playing a role-playing
MUD and gender. The small amount of significant variables nevertheless explains almost 32
percent of the variance in the role-player type.
Table D.3 displays the regression analysis for the griefer component. Again, gender
plays a significant role in grieving with men scoring higher on the griefer factor than women.
Although this seems plain and could be linked to the stereotypical view about men using
violence and women using more covert ways of persuasion and manipulation. The griefer
type as a concept should cover both behaviours, though. Age is the most important
explaining variable, with younger players scoring higher on grieving than older ones.
Stereotypes about the aggressive younger gamer should fit the griefer type. An interesting
predictor is the having of children. Of course this could be moderated by other variables, but
on the whole, people with children score higher on the grieving factor, controlled for all other
variables. For player characteristics, hours spent playing and years playing are both
significant, but have a rather small effect compared to age and gender. The effect of MUD
experience, as measured by years playing, together with age links to the first stage of the
career path mentioned by Bartle (see chapter two, paragraph 2.3). Being more than a mere
player is also significant. So having advanced in the game could lead to more griefing. This
could be linked to having a powerful position with the means to execute griefing in another
way. Of course it would be wrong for administrators and other immortals to abuse their
position.
The type of MUD, either role-playing or player-killing is an interesting one, with playerkilling MUDs scoring high on grieving and role-playing MUDs scoring low on grieving. Playerkilling is very much linked to grief behaviour and the competition on those MUDs leads to
competition and negative feelings to other players. Role-playing MUDs are more about the
story instead of about annoying and imposition upon other players. The fact that so many
demographic and player characteristics are significant is compliant with the player axis of
Bartles model. Players are more important in this respect, with age and gender as the main
predictors, this model explains almost 15 percent of the variance in griefing.
The regression of the socialiser type can be found in table D.4. Gender is again
important, being the second important predictor. Women are more like to be a socialiser than
men. This seems to fit the stereotype, of course. Age is also significant, but barely. Younger
players are a little more sociable than older players. This could be due to younger people
25
The fact that playing an educational MUD relates to role-players significantly, clearly states the importance of this variable,
given the small amount of educational MUDs that were in the sample.
40
being less afraid to make contact with other players. The other demographic variables are
nowhere near significant, although marital status and children could have been good
predictors in theory. The hours spent playing is just barely significant on a two-tailed
significance level of .10, with socialisers spending just a little more time on the MUD. Being
more than just a player predicts a larger amount and is highly significant. This is hardly
surprising when looked at the responsibilities of administrators and immortals towards the
player base. They should be social towards their player base and mostly do not have to
bother with playing themselves. Experience in years or playing multiple MUDs are
insignificant for the socialiser.
The main predictor for the socialiser is the not playing of player-killing MUDs. Of
course, this is not surprising since player-killing is quite the opposite of socialising. The
negative and significant effect of playing educational MUDs is very interesting, because an
educational MUD would be about socialising and helping other players to learn something. It
might be caused by the item about grouping up, since that might not be available on
educational MUDs. The almost significant score of .066 for playing social MUDs (again
probably not significant because of the small amount of associated cases) is not surprising at
all. Social MUDs will cherish socialising. All in all, this component is the least explaining
regression model of the five with an R2 of only 6.7%. The main predictors are not playing a
PK MUD, gender and being more than a mere player.
The last regression, that of the explorer, can be found in table D.5. Women seem less
likely to become an explorer, as are older people. This could maybe be linked to the
stereotypes about women not being technical, since the explorer type has items about bugs
and game mechanics. Also the show off of knowledge is part of the explorer and therefore
more related to the stereotype of the man. Other demographic variables do not seem to
predict very much, given the fact that the betas are rather small and insignificant. The player
characteristics explain more of the variance it seems. Although experience is only significant
on a two-tailed .10 level, the hours spent playing and being more than a mere player are
highly significant. The first relates to the time that is needed to explore the game in all its
facets and the latter relates to the fact that the depth of the MUD and the dealing with bugs
and game mechanics might be more interesting to advanced players like builders, immortals,
administrators and coders. They are concerned with the MUD as a whole, and bugs lead to
trouble.
Only playing role-playing games is significantly related to the explorer. A player of
such a role-playing MUD will less likely be an explorer. This might be due to the fact that
role-playing is about the story and not about the world itself. It does not matter how many
secrets you find or to know the game mechanics. The experience of immersion will be much
41
more important. Overall, the demographic and being more than a mere player predict the
most of the variance in this model, while this model explains just barely over ten percent of
the variation.
4.2.3
Discussion
All in all, the above presents us with the proof that different behaviours link up to specific
player types. Of course, it was a guided research, but the fact that it reproduces the four
player types of Bartle and the additional role-player based upon the immersion motivation of
Yee. One interesting finding is the fact that the explorer comes up, while it was not found in
MMORPGs. This might be linked to the expectation posed in chapter two about the
possibility of advancement within the game, relating to more depth (for instance by builders
and coders) and the ease of creating something with text. As could be seen in the above
multiple regressions, the player types could be explained by demographic variables as well
as MUD and player characteristics. Table 4.5 (next page) is a summary of the main findings
of the regressions. For every component the beta is given and significant betas are marked.
The only demographic variable that is significant for all player types is gender. Men
are more likely to be achievers, griefers or explorers, while women are more likely to be roleplayers or socialisers. Age is significant in almost all player types, except the role-player. It
has the biggest effect on the griefer. An interesting finding is that all betas for age are
negative. This indicates that younger players are more likely to adopt either one of the
playing styles. A possible explanation could be in hours spent playing, which shows the
same significance per player type. A single linear regression of hours spent on age reveals
that this is not the case because the explained variance is just 1% and the coefficient is not
significant. The effects of marital status and having a higher education are never significant
or even come very close. Having children is only significant in the griefer player type. A
possible explanation is the fact that parents become more competitive again when they have
children.
For the player characteristics the main predictors are being more than a mere player
and the hours spent playing. Being more than a mere player influences the player style, but
only achievers are more likely to be mere players. The effect of being more than a mere
player is strongest in the explorer. All player types spent more hours in the game, but this is
strongest for the explorer as well. Years playing is only significant for the explorer and the
griefer. For the explorer this is not surprising, because a newbie will hardly start to explore
the width and breadth of the MUD. A newbie will be more concerned with learning the basics.
The significance of experience for griefers might stem from an air of I am powerful, because
I have been here forever and therefore I can bully all other players. This seems quite the
opposite from the huge negative age effect for griefers. The young, but experienced players
42
seem to be very likely to impose themselves on their fellow players. Playing more than one
MUD seems not to affect any player type.
Table 4.5: Summary table of the s for the achiever, role-player, griefer, socialiser and explorer,
based upon the tables D.1-D.5.
Achiever
Variables
Role-player
2
Griefer
2
Socialiser
2
Explorer
2
R =.111
R =.319
R =.145
R =.067
R =.105
-.226 (.000)
.163 (.000)
-.194 (.000)
.112 (.000)
-.156 (.000)
- age (Years)
-.166 (.000)
-.044 (.087)
-.309 (.000)
-.068 (.024)
-.112 (.000)
.012 (.672)
-.037 (.125)
-.016 (.541)
.015 (.602)
.020 (.458)
.051 (.082)
.015 (.564)
.080 (.005)
.029 (.328)
-.025 (.389)
.010 (.705)
-.011 (.617)
.014 (.578)
-.007 (.776)
-.019 (.451)
Demographic Variables
Player Characteristics
- hours spent playing (Hours)
.058 (.021)
.032 (.139)
.056 (.023)
.053 (.036)
.084 (.001)
.036 (.182)
-.037 (.119)
.089 (.001)
.010 (.721)
.054 (.049)
.040 (.111)
.002 (.919)
.005 (.846)
.019 (.454)
-.002 (.944)
-.107 (.000)
.080 (.000)
.053 (.034)
.108 (.000)
.144 (.000)
-.134 (.000)
.490 (.000)
-.105 (.000)
-.022 (.460)
-.111 (.000)
.127 (.000)
.038 (.134)
.135 (.000)
-.145 (.000)
-.022 (.457)
-.057 (.077)
.040 (.155)
-.005 (.881)
.060 (.066)
-.004 (.893)
.011 (.722)
-.070 (.012)
.015 (.621)
-.079 (.016)
.004 (.891)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
The type of MUD seems to predict several of the player types. Role-playing MUDs
and player-killing MUDs are the most predictive, but this is hardly surprising given the small
amount of educational and social MUDs in the sample. Nevertheless playing an educational
MUD is negatively related to both the role-player and the socialiser and playing a social MUD
comes close to significance on the achiever (negatively) and the socialiser. Educational
MUDs do not attract role-players it seems, which is not surprising given the fact that
educational is rooted in real life and role-playing is about immersion in another world. The
fact that playing a social MUD predicts the socialiser, but playing an educational MUD does
not, is probably due to the fact that social MUDs are about socialising to begin with, while
playing educational MUDs are about personal education and not related to socialising.
Playing a role-playing MUD predicts being a role-player, which is hardly surprising, all other
player types relate negatively with the role-player. Role-playing MUDs might harbour different
player types, but those do not correspond much to those of Bartle. Playing a player-killing
MUD predicts the griefer and achiever positively and the socialiser negatively. This is hardly
43
For matters of ease openness will be used from here onwards for the intellect, openness, or imagination factor.
44
doing inside their head from a purely text-based input. A critical note must be made about
this, though, because other variables (for instance, the high amount of highly educated
respondents) could be related to this skewedness.
By doing a quick principal component analysis to validate if the factors were all
related as intended, the Eigen value > 1 brought nine items, but the scree plot indicated
clearly that five factors were standing out. After doing a varimax and oblimin rotation with five
factors the simple structure advocated by Saucier (1994), became very much apparent27 in
both. Therefore it seems that the intended scale was reproduced and can be used for a sum
score, which was constructed by adding the items. The reliability of the scale was also tested
and the results can be found in table 4.6. When the sum scores were displayed in a plot, the
earlier conclusion that agreeableness and openness would score high was confirmed,
because both were skewed towards the right. The other factors were normally distributed on
the scores.
Table 4.6: The Big Five Factors, associated items and their reliability as a scale.
28
Reliability
Positive Items
Negative Items
Extraversion
= .826
Talkative, Extraverted,
Bold, Energetic
Agreeableness
= .844
Sympathetic, Warm,
Kind, Cooperative
Cold, Unsympathetic
Rude, Harsh
Conscientiousness
= .822
Organized, Efficient,
Systematic, Practical
Emotional Stability
= .805
Disorganized, Sloppy,
Inefficient, Careless
Moody, Jealous,
Temperamental, Envious,
Touchy, Fretful
Intellect,
Openness,
or Imagination
= .792
29
Unenvious, Relaxed
Creative, Imaginative,
Philosophical, Intellectual,
Complex, Deep
Uncreative, Unintellectual
Because personality is inherent to the person and when we take into account that
personality, or at least the extraversion and emotional stability factors, are fairly stable as
was discussed in chapter two, a multiple regression would not give us much related
information. Therefore the factors will be correlated only to gender and age (which are also
mostly used as independent variables), before we link them to the online playing styles.
27
By counting the items per factor one factor with 9 items and one with only 7 items was found. When determining where this
came from, it was discovered that there had been a typing error in the survey: SHY had been written as SLY. Since this
word also exists in English, this was not discovered before the test started. Although the item loaded quite strong on the
factor of agreeableness, it was deemed wiser to remove it from the analysis. Therefore there are only 39 items in the
results, with only seven instead of eight factors for extraversion (missing shy).
28
Reliability based on Cronbachs alpha (total item-correlation > 0.3 and alpha if item deleted < scale alpha)
29
Reliability analysis indicated that if relaxed was removed it would be .806. For matters of keeping Sauciers scale in tact and
the small increase the item was left in the scale.
45
When we correlate gender with the factors, only extraversion shows no significant
difference between the means for men and women (p=.875). On the whole, women score
significantly higher on agreeableness, openness (both p<.001) and conscientiousness
(although only significant on .10 for the latter, two tailed.) If we correlate the factors with age,
extraversion (r = .006) and openness (r = 0.10) correlate barely. The correlations between
the other three are all positively and significantly correlated, but also with a low correlation
(agreeableness, r = .193, conscientiousness, r = .145, emotional stability, r = .109).
4.3.2 Offline Traits Used to Explain Online Player Types
Now it is time to see if the offline personality traits can be used in explaining online
behaviour. Therefore we will revisit the multiple regression analyses from paragraph 4.2.2
and add the five factors. Then we will determine if this leads to a significant increase in
explained variance and therefore be a better model. To save up space, these multiple
regression tables can be found in appendix D.
The regression with the big five factors added for the achiever can be found in table
D.6. Achievement seems to be linked to almost all personality traits, except openness. The
rest of the regression table is quite similar, but for the hours spent playing, which lost its
significance when the personality traits were added. Personality might explain some of the
variance in playing time. A regression of hours spent playing on the five personality traits
reveals barely two-tailed .10 significance (p<.050) for conscientiousness, emotional stability
(both negatively related) and openness (positively related). Gender stays the main predictor,
but
conscientiousness
and
emotional
stability
are
second
and
third
in
place.
46
The regression analysis for the role-player is in table D.7. This picture provides less
significant big five factors than the achiever regression did. Only conscientiousness and
openness are significant, with the openness factor being among the most important
predictors. This was also expected in chapter two, because role-playing involves the
imagination and openness to experience to be able to be the character. The
conscientiousness factor is also important and this could be linked to self-restraining. The
goal is to stay in character and being efficient and doing this systematic will probably work
better than if you would play in a careless manner. The fact that extraversion is not
significant, hints that the link that Bartle placed between socialising and role-playing is noneexistent. Emotional stability and agreeableness are not significant either, although agreeing
with other players or at least with their idea of the story or character could have been.
Emotional stability is another issue, for fleeing into an online character or story could save
some real life stress. But the insignificance of it clearly supports Yees finding of two separate
motivations for immersion (role-playing) and escapism (fleeing real life). An interesting
variable is years playing, since it is significant with the personality added in, while it was not
before. The more years you have been playing, the less likely you are to be a role-player.
This model also explains more variance than the one without the personality factors and it
goes up from the already impressive 31.9% to 36.2%.
Next is the regression of the griefer (see table D.8). The most notable change is the
explained variance of course, which has nearly doubled by adding the big five factors. This is
not really surprising, since the killers business is annoying people, and although they do it in
the game, it is based upon real life. The more statistical explanation is that there are four Big
Five factors with a significant influence. Looking at the standardized betas negative
agreeableness seems to be the main predictor in this. Of course this is obvious, because
annoying players and being rude and unsympathetic is the way of the killer. Next to gender
and age, extraversion is also an important predictor and again this is no surprise, since you
need to dare to talk to players and people and be just plain bold on the whole to be able to
create havoc and grief. Openness is related to this as well. Bullying people in the same way
every time will be boring and being inventive about it might relate to fun. Emotional stability is
negatively linked to the griefer player type. So being less emotionally stable will result in
more griefer behaviour. This could be about a reaction on negative experiences in real life.
People might be compensating the negativity of real life by bullying others when they are
online. Another interesting finding is that the player characteristics are less explaining, when
the five personality traits are added to the model. Hours spent playing and being more than a
mere player are not significant any longer. The exact reason for this can only be guessed.
The socialiser also relates a lot to the personality traits as can be seen in the
regression analysis table D.9. The adding of the personality factors lead to an increase of
47
over thirteen percent, making a very good predictive model for the social sciences. Even all
the Big Five factors are significant. Extraversion and agreeableness seem logical traits to
have when being a socialiser and this is confirmed with them being the main predictors in the
model. An open mind also comes in handy when dealing with players that you do not know
well yet. Conscientiousness and emotional stability are both negatively related to the
socialiser, which is not really surprising. Emotionally unstable persons want reassurance and
socialising could be the way. Also the goal-oriented view of conscientious people could leave
less room for mere socialising, leaving the players that score lower on the trait to be more
socially oriented. Playing an educational MUD and age have lost their significance in this
newer model. This could be congruent with players of educational MUDs playing for their
own personal incentives rather than not wanting to socialise. That age lost its significance
might be due that socialising is of all ages, controlling for personal traits.
The last player type is the explorer which can be found in table D.10. In this
regression years playing has lost its small significance. The main explaining factors seem to
be gender (men are more likely to be explorer) and the additional factor of openness (higher
scores are more likely to be explorers). This last is not surprising. The name associated with
high scores on openness was explorer and an open mind is the way to explore. Exploring is
about the discovery of new areas and new content. You have to be open to new
experiences. Almost all big five factors are significant (except for extraversion). Next to the
above mentioned openness, conscientiousness is positively related. A systematic way of
exploring might be the most successful. Emotional stability and agreeableness are negatively
related. The first might be related to the attraction of negative experiences that a lower
emotional stability might cause. Exploring might be highly trial and error and negative
experiences just happen that way. A lower score on agreeableness leading to a higher score
on exploring is hardly surprising though, because doing things differently than others, just to
see what it will bring, needs people that do not rely on the standards that are supplied by the
masses.
After having discussed the regression models with the additional personality traits it
would be good to see if the changes in R2 are significant. If they are not, the risen predictive
quality might be due to chance or sampling errors. In order to check for this, five F statistics
were calculated comparing the initial models with the improved models and were checked for
significance. All turned out to be significantly (=.01) improved models (Fachiever = 11.082,
Frole-player = 18.211, Fgriefer = 52.906, Fsocialiser = 44.725, and Fexplorer = 12.641)
4.3.3
Discussion
The above regression models showed that adding the personality factors of the Big
Five lead to an increased predictive quality with higher explained variance. In addition to the
48
demographic variables and the MUD and player characteristics, playing styles seem to relate
to the offline personality. This enables the conclusion that the player types are in a certain
way related to the offline characteristics of the person. As can be seen in table 4.7, listing the
increases in explained variance, especially the player oriented player types of Bartles
interest graph (griefer and socialiser) were much better explained by adding the personality
traits. The Big Five regression coefficients together with their significance are summarised in
table 4.8. Significant effects are marked yellow.
Table 4.7: Increases of the explained variances for the initial models and the additional models.
Initial R
Achiever
Role-Player
Griefer
Socialiser
Explorer
R =.111
2
R =.319
2
R =.145
2
R =.067
2
R =.105
R =.146
2
R =.362
2
R =.285
2
R =.199
2
R =.145
Increase
4.5 %
4.3 %
14.0 %
13.2 %
4.0 %
Extraversion is linked to the socialiser and griefer, which are not surprising given their
player component. Next to this it was also related to the achiever. This might be related to
Hedron's prove mastery circle in which players want to show their mastering of the game by
either helping (socialiser) or killing (griefer). Agreeableness was skewed a little to the right to
begin with. Nevertheless it was significant in four of the five player types, with it having a
negative effect on the achiever, griefer and explorer and having a positive effect on the
socialiser. These are not surprising. A socialiser needs agreeableness in order to keep the
peace between his/her friends. Griefing is exactly the opposite and with achievers striving to
reach their goals, they will have to follow their own path most of the time. Explorers are
interested in the game mechanics and such and think they know how it works. This attitude
may not be appreciated by all.
Table 4.8: Summary table of the personality trait s for the achiever, role-player, griefer, socialiser
and explorer, based upon the tables D.6-D.10.
Variables
Achiever
Role-player
Griefer
Socialiser
Explorer
.094 (.000)
.027 (.239)
.173 (.000)
.218 (.000)
-.032 (.230)
-.055 (.048)
.022 (.366)
-.347 (.000)
.260 (.000)
-.069 (.013)
.151 (.000)
-.060 (.010)
-.006 (.808)
-.080 (.002)
.064 (.017)
-.150 (.000)
-.008 (.722)
-.078 (.002)
-.104 (.000)
-.059 (.030)
.023 (.389)
.202 (.000)
.086 (.000)
.073 (.004)
.165 (.000)
49
Conscientiousness, being related to goals and keeping an eye on the overall pattern
should relate to the achiever and explorer. In that it does hold, since both have a significant
effect according to the regression analysis. The fact that the socialiser is negatively related to
conscientiousness is probably because socialising is less goal-oriented. The only surprise
might be the role-player that is negatively related to conscientiousness. This could be due to
the fact that when you cannot hold the overall view you immerse yourself more in the game
and your character. Emotional Stability is negatively linked to all player types, only with the
role-player it is not significant. As mentioned in chapter two, negative experiences happen to
anyone and not just to a single player type. Only maybe the victims of killers, but those could
be of any player type. Why it has no significant effect on role-players, remains the question.
Maybe this is due to their view that they play in a story and can relate the negative feelings
towards the story instead of themselves. If this is the case should be seen in further
research.
The last big five factor was the most important one for MUD players. At least it was
thought so in chapter two. The overall distribution was skewed to the right to begin with,
supporting that you need a highly creative and open mind to immerse yourself in the textbased game world. The regression analyses for the player types reveal that it plays an effect
in almost all player types. Only achievers are not predicted significantly by openness to
experience. This might well be a side-effect of the achievement motivation. When you are
trying to gain status you will use the known and sure ways and you will not try out the new
and uncertain ones, unless the rewards could be extraordinary.
Overall, the player types could be found and they could be explained by using
demographic variables, MUD and player characteristics and the Big Five personality traits.
The links between the type of MUD and the player types seem very much apparent and
therefore type of MUD might have more influence in which player types inhabit a specific
MUD than that different player types can be found on any MUD, like Bartle proposes. In the
concluding chapter a new model will be proposed based upon the above findings.
50
At the end of the thesis it is time see what can be learned from all of this. This chapter will
start with a summary of the findings after which the research question will be answered. After
that a new model for player types will be presented and the chapter ends with
recommendations for further research.
5.1
This research was heavily guided by Bartles typology of the four playing styles and
subsequent research that criticised it. It set out to reveal which types of behaviour could be
clustered together to form specific player types, taking the concepts of the achiever, the
explorer, the killer and the socialiser player types by Bartle and the achievement, the
immersion, the manipulation and the relationship motivations from Yee as a starting point.
Thirty six items were created that should measure the five anticipated player types: the
achiever, the role-player, the explorer, the griefer and the socialiser.
The principal component analysis revealed that five clusters of items could account
for more than fifty percent of the variance among the items, and surprisingly fit the five
anticipated playing styles. The achiever matched with five of the intended items and also with
two additional items: one that was supposed to match the explorer (finding new ways to
level) and one supposedly linked to the griefer type (competition). The griefer also linked all
items nicely to it. For the anticipated socialiser player types, all six intended items fell into the
socialiser category, but helping others was also somewhat negatively linked to the griefer
player type. The additional role-player playing style items were neatly clustered after removal
of the low-correlation items I like to try out new roles and I think that realism in the game is
important. This seems to indicate that Yee was right in assuming that there should be a roleplaying motivation, or at least player type. The explorer was found as well, reproducing six of
51
the initial nine items, after removing two very low correlating items. Surprisingly those were I
like to explore all the areas in the game and I like maps, charts and tables with information
about the game, which seem very basic items for the explorer. It seems that the explorer is
a more complex character, which is also closely linked to the achiever motivation. Bartle
indicated that the explorer would start exploring the breadth of the MUD (the map and areas)
before moving on to its depth (the game mechanics and bugs in the game play). It seems
that the breadth of the MUD does not concern the explorer types and that depth is much
more important.
The second part of the research was about colouring the picture of the different player
types by doing five multiple regressions. It was found that demographic variables together
with player and MUD characteristics explained a great deal of the player types already. One
of the main findings was that type of MUD relates heavily to the type of player with roleplayers playing role-playing MUDs, griefers playing player-killing MUDs, and socialisers
playing social and educational MUDs. Gender was also a significant predictor of the player
types with men being achievers, griefers and explorers and women being role-players and
socialisers. Being more than a mere player was also related to all the player types, although
achievement did it in a negative way, leaving achievers being mere players and the other
player types being more than players.
Next to these explaining variables, it was hypothesized in the conceptual model that
personality traits would explain great deal of the player types. Next to the fact that the
Sauciers mini-marker scale was proven to be robust, it was proven that entering the Big Five
personality factors in the regression analyses lead to significant better explained variances.
Nineteen of the twenty-five (five player types times five personality traits) predictors were
significant. All models were improved by at least four percent explained variance with the
socialiser and griefer models being enhanced with 13.2% and 14.0% respectively. No
personality trait explained every player type, but the socialiser was explained by all five traits
while the role-player was only explained by two traits. Nevertheless the latter already had
more than 30% explained variance in the first model.
player as interacting with the world (or story). Now that this thesis finds evidence for both, it
becomes a lot more difficult. The explorer player type as found in this research has virtually
no correlation with the role-player. In fact the role-player has the highest (but still very small:
.163) correlation with the socialiser type, which links back to Bartle placing the role-playing
component under socialiser. Nevertheless, the lack of significance in extraversion for the
role-playing could indicate the other opposite. So should we put the role-player in between
the socialiser and explorer on the interaction axis? This seems not to make sense, even
more if we look at the other correlations between the components as well. As mentioned
above the explorer and socialiser are both correlated to the achiever. So should we move the
socialiser up in the model, higher on the acting axis? Or should we take it away from the
player end of the axis? Or does the achiever have less to do with the game, should we
center that one? This last line of thought makes even the most sense, since achieving is
different per game type.
When we look at the explaining variables for the player types we gain even more
ground for this approach, since the type of the game that players tend to play seems to work
out in the playing styles displayed. Whether the game influences the player or vice versa
remains to be seen, but the fact is that role-playing is greatly explained by playing a roleplaying MUD, that grieving is explained by using a player-killing MUD, and although the
number of social MUDs has been small in the sample, there was some evidence that playing
a social MUD predicted socialiser. So we have three game types and player types that seem
highly linked. Then we have an achiever that wants to achieve whatever the game goal is
and then we have the explorer, which wants to have an in-depth knowledge of the world.
This also matches with the findings that being more than a mere player is positively linked to
all player types, except the achiever. If we try to put this into a model, it would look like the
one displayed in figure 5.1.
Figure 5.1: A new model for the player types
E X P L O R E R
/
/
A C H I E V E R
/
/
|
|
\
\
SOCIALISER
ROLE-PLAYER
53
If seen in this way, it shows similarity to the hierarchically ordered player types of
Hedron, with the beating the game/excel and proving mastery circles relating to the
achiever player type and the seeking new challenges and all is one circles relating to the
explorer. The lower circles of survival and competence could be linked to the different types
of games, with the newbies adapting to the game or newbies looking for the game that they
want, before they start playing for real. This lends more strength to the hierarchical approach
in general as well.
It seems that the kind of game is very much providing the base of this model, instead
of specific player types. This allows for a comparison between this and the interest graph of
Bartle presented in chapter two. At first glance this interest graph has some similarities with
the realms of experience from Pine and Gilmore (1999, p.30). Pine and Gilmore argue that
the more realms targeted the richer the experience. This is analogous to Bartle promoting the
pursuing of a balance between the player types, more over if we take the above model in
account. If the base provides the realms of fun for Bartle, the combination of role-playing,
player-killing and socialising might provide a rich game. Also a rich game would provide more
interesting options for players that have reached the seeking new challenges stage of
Hedron and start exploring the full scope of the MUD as explorers. This would prevent them
from choosing the eighth option: finding another game. More research about the type of
games should be conducted to verify this claim, but it sounds plausible.
The base of the model with different MUDs lead to achievers wanting to advance in
the game and every type of MUD provides different goals, but achievement is possible in
every game. As Clodius (1994, in Bartle,1996) claims: Social MUDs do have their achievers,
too: people who regard building as a competitive act, and can vie to have the best rooms in
the MUD. Finally, players will reach the state of the explorer, in which people are
experimenting with aspects of the game that they did not know yet. The model provides a
rich base full of players and just a smaller and smaller cone reaching up towards the
explorers. This also fits Bartles notion that explorers are hard to find and if you have them,
you need to keep a hold on them. Combined with Bartles criticism on the commercial MUDs
and MMOGs that fail to invest in the dedicated player base, this might explain why Yee did
not find the explorer. In MMOGs they have a very large player base, but the opportunities to
make a career are mostly not provided, leaving only the last option of Hedrons challenges
circle: finding another game.
54
like the experience economy. The real challenge, however, will lie in proving or disproving
this new model. Furthermore the link with offline personality traits will need to be investigated
further. Although they predict the playing styles, the exact reason and origin of the links
between some traits and player types is hard to make out by the data provided in this study.
Subsequent research should concern itself with validating the presented player types model
and causal analysis of the links between offline traits and online behaviour. This thesis can
therefore be used as a guide.
56
6. References
6.1 Books/Articles
Aarseth, E. J. (2003) Playing Research: Methodological approaches to game analysis,
Melbourne DAC - the 5th International Digital Arts and Culture Conference, School of
Applied
Communication,
Melbourne.
Retrieved
August
1,
2007,
from
http://spilforskning.dk/gameapproaches/GameApproaches2.pdf.
Bartle, R. (1996). Hearts, Clubs, Diamonds and Spades: Players who suit MUDs. Journal of
Online
Environments,
(1).
Retrieved
January
11,
2007,
from
http://www.brandeis.edu/pubs/jove/HTML/v1/bartle.html.
Bartle, R. (2003). Designing Virtual Worlds, Berkeley, CA: New Riders Publishing
Blossom, J. (2006, October 23). Get a Life: Second Life Points the Way Towards Content
Growth
in
Real-World
Communities.
Retrieved
December
17,
2006,
from
http://shore.com/commentary/newsanal/items/2006/20061023secondlife.html.
Dibbell, J. (1998) My Tiny Life: Crime and passion in a virtual world, New York, NY: Henry
Holt. Retrieved June 12, 2007, from http://www.juliandibbell.com/texts/bungle.html.
Farmer, F.R. (1992). Social Dimensions of Habitats Citizenry. In Loeffler, C.E. & Anderson,
T. (Eds). The Virtual Reality Casebook, New York: Van Nostrand Reinhold. Retrieved,
June 9, 2007, from http://scholar.google.nl/scholar?hl=nl&lr=&q=cache:WwIkxKUAjYYJ:
sunsite.tus.ac.jp/pub/academic/communications/papers/habitat/citizenry.rtf.
Foo, C.Y. & Koivisto, E.M.I. (2004) Grief Player Motivations, paper presented at the Other
Players conference, Center for Computer Games Research, IT University of
Copenhagen, Denmark.
Fullerton, T., Swain, C. & Hoffman, S. (2004) Designing, Prototyping, and Playtesting
Games. San Francisco, CA: CMP Books.
57
5,
2007, from
http://eric.ed.gov/
ERICDocs/data/ericdocs2/content_storage_01/0000000b/80/27/07/32.pdf.
Keegan, M. (2003) A Classification of MUDs, Journal of Online Environments, 2 (2). Updated
version retrieved June 6, 2007, from http://mk.ucant.org/info/classification_muds.html.
Lazzaro, N. (2004). Why We Play Games: four keys to more emotion without story. Retrieved
June 9, 2007, from http://xeodesign.com/xeodesign_whyweplaygames.pdf.
McAdams, D. P. (1994). The Person: An introduction to personality psychology. Fort Worth:
Harcourt Brace College Publishers
Oates, J. (2006, June 23). Name a website with 22bn page impressions? Retrieved
December 17, 2006, from http://www.theregister.co.uk/2006/06/23/cyworld_comes_
to_germany.
Pallant, J. (2004). SPSS Survival Manual, Buckingham: Open University Press.
Pine II, B.J. & Gilmore, J. H. (1999) The Experience Economy: Work is theater & every
business a stage. Boston: Harvard Business School Press.
Reid, E. M. (1994). Cultural Formations in Text-Based Virtual Realities, Master Thesis
Cultural Studies, Melbourne: University of Melbourne. Retrieved June 5, 2007, from
http://www.aluluei.com/cult-form.htm.
Rheingold, H. (2000). The Virtual Community: homesteading on the electronic frontier.
Cambridge, Massachusetts: MIT Press. Retrieved December 16, 2006, from
http://www.rheingold.com/vc/book/index.html.
58
12,
2007,
from
ftp://jefferson.village.virginia.edu/pub/pubs/pmc/pmc-
moo/MOO.Terrorism.
Yee, N. (2004, September 21). Unmasking the Avatar: The demographics of MMO player
motivations, in-game preferences, and attrition. Gamasutra. Retrieved June 12, 2007,
from http://m-plabs.com/mmogarticles/MMO_Player_Motivations.pdf.
Yee, N. (2006a). The Demographics, Motivations and Derived Experiences of Users of
Massively-Multiuser Online Graphical Environments. Presence: Teleoperators and
Virtual Environments, 15 (3), p. 309-329. Retrieved January 9, 2007, from
http://www.nickyee.com/pubs/Yee%20-%20MMORPG%20Demographics%202006.pdf.
Yee, N. (2006b). Motivations for Play in Online Games, CyberPsychology & Behavior, 9 (6),
p. 772-775.
6.2 Websites
www.amazon.com
Amazon.com
www.andreasen.org/bartle
www.british-legends.com
British Legends
www.guildcafe.com
www.guildcafe.com/bartle.php
www.mudconnect.com
http://www.mudconnect.com/mpd/
www.secondlife.com
Second Life
59
Explorer
Socialiser
Role-Player
Killer
-,734
,716
,691
,673
,561
,551
,827
,776
,759
-,717
-,677
,662
,598
-,848
-,809
-,704
-,661
,661
,393
-,397
,746
,644
,606
-,600
,572
,367
30
,383
,702
,695
,657
,653
,628
,557
30
This score was left out of the griefer scale, because Cronbach's alpha would be higher if the item was deleted.
Table D.1: Multiple Regression for the achiever component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD and educational MUD.
Model
SD
(constant)
25.465
.603
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-2.940
-.105
.133
.671
.119
.331
.019
.315
.386
.313
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
.022
.052
.501
-1.305
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
-1.750
1.479
-3.164
.726
Beta
Sig
42.249
.000
-.226
-.166
.012
.051
.010
-8.883
-5.564
.423
1.738
.379
.000
.000
.672
.082
.705
.009
.039
.315
.315
.058
.036
.040
-.107
2.318
1.335
1.592
-4.143
.021
.182
.111
.000
.371
.340
1.785
2.036
-.134
.127
-.057
.011
-4.714
4.344
-1.772
.357
.000
.000
.077
.722
R =.111
Table D.2: Multiple Regression for the role-player component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD and educational MUD.
Model
SD
21.282
.532
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
2.142
-.029
-.427
.197
-.138
.292
.017
.278
.341
.276
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
.012
-.054
.028
.993
6.451
.452
2.244
-4.542
(constant)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Beta
Sig
40.005
.000
.163
-.044
-.037
.015
-.011
7.346
-1.710
-1.537
.577
-.500
.000
.087
.125
.564
.617
.008
.035
.277
.277
.032
-.037
.002
.080
1.480
-1.562
.101
3.581
.139
.119
.919
.000
.329
.301
1.577
1.798
.490
.038
.040
-.070
19.629
1.500
1.423
-2.526
.000
.134
.155
.012
R =.319
Table D.3: Multiple Regression for the griefer component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD and educational MUD.
Model
SD
(constant)
16.829
.502
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-2.150
-.168
-.161
.904
.145
.276
.016
.262
.322
.261
.018
.110
.051
.556
-1.170
1.337
-.224
.842
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Beta
Sig
33.513
.000
-.194
-.309
-.016
.080
.014
-7.774
-10.629
-.612
2.806
.556
.000
.000
.541
.005
.578
.008
.033
.262
.263
.056
.089
.005
.053
2.283
3.370
.195
2.119
.023
.001
.846
.034
.311
.285
1.491
1.701
-.105
.135
-.005
.015
-3.767
4.693
-.150
.495
.000
.000
.881
.621
R =.145
Table D.4: Multiple Regression for the socialiser component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD and educational MUD.
Model
(constant)
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
SD
22.613
.403
.950
-.029
.110
.252
-.059
.221
.013
.210
.258
.209
.013
.009
.157
.860
-.184
-1.101
2.197
-3.292
Beta
Sig
56.055
.000
.112
-.068
.015
.029
-.007
4.295
-2.255
.522
.978
-.284
.000
.024
.602
.328
.776
.006
.026
.210
.210
.053
.010
.019
.108
2.094
.357
.749
-4.094
.036
.721
.454
.000
.248
.228
1.196
1.364
-.022
-.145
.060
-.079
-.740
-4.835
1.838
-2.414
.460
.000
.066
.016
R =.067
Table D.5: Multiple Regression for the explorer component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD and educational MUD.
Model
SD
(constant)
22.076
.459
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-1.545
-.054
.179
-.254
-.179
.252
.014
.241
.294
.238
.024
.059
-.017
1.344
-1.103
-.193
-.183
.213
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Beta
Sig
48.068
.000
-.156
-.112
.020
-.025
-.019
-6.135
-3.749
.742
-.861
-.753
.000
.000
.458
.389
.451
.007
.030
.240
.240
.084
.054
-.002
.144
3.369
1.974
-.071
5.610
.001
.049
.944
.000
.283
.260
1.361
1.552
-.111
-.022
-.004
.004
-3.900
-.744
-.134
.137
.000
.457
.893
.891
R =.105
Table D.6: Multiple Regression for the achiever component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD, educational MUD, extraversion, agreeableness,
conscientiousness, emotional stability and intellect, openness, or imagination.
Model
SD
(constant)
24.280
1.429
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-3.073
-.088
-.212
.610
-.115
.353
.020
.328
.397
.324
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
.015
.044
.448
-1.200
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Big Five Personality Traits
- extraversion
- agreeableness
- conscientiousness
- emotional stability
- intellect, openness, or imagination
2
Beta
Sig
16.993
.000
-.238
-.137
-.019
.047
-.009
-8.712
-4.362
-.646
1.538
-.355
.000
.000
.518
.124
.723
.020
.040
.326
.330
.041
.031
.036
-.098
1.565
1.083
1.375
-3.634
.118
.279
.169
.000
-1.697
1.137
-3.099
1.270
.390
.355
1.757
2.026
-.131
.099
-.058
.020
-4.357
3.199
-1.764
.627
.000
.001
.078
.531
.052
-.031
.079
-.078
.016
.015
.015
.014
.014
.019
.094
-.055
.151
-.150
.023
3.573
-1.979
5.634
-5.516
.862
.000
.048
.000
.000
.389
Table D.7: Multiple Regression for the role-player component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD, educational MUD, extraversion, agreeableness,
conscientiousness, emotional stability and intellect, openness, or imagination.
Model
SD
14.145
1.259
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
1.756
-.021
-.357
.128
-.281
.310
.018
.289
.394
.285
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
.012
-.096
-.114
.703
(constant)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Big Five Personality Traits
- extraversion
- agreeableness
- conscientiousness
- emotional stability
- intellect, openness, or imagination
2
Beta
Sig
11.237
.000
.134
-.032
-.031
.010
-.022
5.657
-1.168
-1.235
.366
-.988
.000
.243
.217
.714
.323
.009
.036
.287
.290
.032
-.066
-.009
.057
1.424
-2.691
-.396
2.424
.155
.007
.692
.015
6.125
.465
1.879
-5.176
.344
.314
1.547
1.784
.463
.040
.034
-.081
17.810
1.482
1.214
-2.901
.000
.139
.225
.004
.015
.012
-.032
-.004
.145
.013
.014
.012
.013
.016
.027
.022
-.060
-.008
.202
1.179
.904
-2.596
-.356
8.815
.239
.366
.010
.722
.000
Table D.8: Multiple Regression for the griefer component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD, educational MUD, extraversion, agreeableness,
conscientiousness, emotional stability and intellect, openness, or imagination.
Model
SD
(constant)
20.341
1.116
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-1.848
-.122
-.241
.977
-.075
.275
.016
.256
.309
.252
.010
.094
.080
.459
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Big Five Personality Traits
- extraversion
- agreeableness
- conscientiousness
- emotional stability
- intellect, openness, or imagination
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
Beta
Sig
18.232
.000
-.168
-.224
-.025
.088
-.007
-6.711
-7.809
.941
3.164
-.297
.000
.000
.347
.002
.767
.008
.032
.255
.257
.033
.077
.008
.044
1.369
2.965
.315
1.785
.171
.003
.753
.074
-.936
1.245
-.164
.930
.305
.278
1.371
1.581
-.085
.126
-.004
.017
-3.069
4.474
-.120
.589
.002
.000
.905
.556
.082
-.166
-.003
-.035
.052
.011
.012
.011
.011
.015
.173
-.347
-.006
-.078
.086
7.153
-13.749
-.244
-3.138
3.547
.000
.000
.808
.002
.000
Table D.9: Multiple Regression for the socialiser component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD, educational MUD, extraversion, agreeableness,
conscientiousness, emotional stability and intellect, openness, or imagination.
Model
SD
16.009
.918
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
.540
-.024
-.261
.238
-.258
.226
.013
.210
.254
.208
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
.017
.010
-.082
.722
(constant)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Big Five Personality Traits
- extraversion
- agreeableness
- conscientiousness
- emotional stability
- intellect, openness, or imagination
2
Beta
Sig
17.442
.000
.063
-.056
-.035
.027
-.031
2.385
-1.865
-1.243
.937
-1.243
.017
.062
.214
.349
.214
.006
.026
.210
.212
.070
.011
-.010
.089
2.773
.398
-.390
3.411
.006
.691
.697
.001
-.368
-1.371
2.173
-1.666
.251
.228
1.130
1.303
-.043
-.179
.061
-.040
-1.470
-6.003
1.922
-1.279
.142
.000
.055
.201
.081
.097
-.028
-.036
.034
.009
.010
.009
.009
.012
.218
.260
-.080
-.104
.073
8.563
9.756
-3.080
-3.932
2.857
.000
.000
.002
.000
.004
Table D.10: Multiple Regression for the explorer component on gender, age, marital status, children,
higher education, hours spent playing, years playing, multiple MUDs, more than mere player, roleplaying MUD, player-killing MUD , social MUD, educational MUD, extraversion, agreeableness,
conscientiousness, emotional stability and intellect, openness, or imagination.
Model
SD
(constant)
19.109
1.078
Demographic Variables
- gender (0=male, 1=female)
- age (Years)
- marital status (0=single, 1=together)
- children (0=no, 1=yes)
- higher education (0=no, 1=yes)
-1.662
-.039
.127
-.242
-.297
.266
.015
.248
.300
.244
.020
.039
-.050
1.245
Player Characteristics
- hours spent playing (Hours)
- years playing (Years)
- multiple MUDs (0=no, 1=yes)
- more than mere player (0=no, 1=yes)
MUD Characteristics
- role-playing MUD (0=no, 1=yes)
- player-killing MUD (0=no, 1=yes)
- social MUD (0=no, 1=yes)
- educational MUD(0=no, 1=yes)
Big Five Personality Traits
- extraversion
- agreeableness
- conscientiousness
- emotional stability
- intellect, openness, or imagination
2
Beta
Sig
17.728
.000
-.171
-.081
.015
-.025
-.032
-6.246
-2.581
.511
-.809
-1.216
.000
.010
.609
.419
.224
.007
.031
.247
.249
.071
.036
-.005
.135
2.744
1.261
-.203
4.997
.006
.207
.839
.000
-1.357
-.197
.189
.941
.295
.269
1.326
1.529
-.139
-.023
.005
.020
-4.608
-.733
.142
.615
.000
.464
.887
.538
-.013
-.029
.025
-.023
.088
.011
.012
.011
.011
.014
-.032
-.069
.064
-.059
.165
-1.202
-2.500
2.395
-2.173
6.219
.230
.013
.017
.030
.000